Ring artifacts are prevalent in 3D cone-beam computed tomography (CBCT) due to non-ideal responses of X-ray detectors, substantially affecting image quality and diagnostic reliability. Existing state-of-the-art (SOTA) ring artifact reduction (RAR) methods rely on supervised learning with large-scale paired CT datasets. While effective in-domain, supervised methods tend to struggle to fully capture the physical characteristics of ring artifacts, leading to pronounced performance drops in complex real-world acquisitions. Moreover, their scalability to 3D CBCT is limited by high memory demands. In this work, we propose Riner, a new unsupervised RAR method. Based on a theoretical analysis of ring artifact formation, we reformulate RAR as a multi-parameter inverse problem, where the non-ideal responses of X-ray detectors are parameterized as solvable physical variables. Using a new differentiable forward model, Riner can jointly learn the implicit neural representation of artifact-free images and estimate the physical parameters directly from CT measurements, without external training data. Additionally, Riner is memory-friendly due to its ray-based optimization, enhancing its usability in large-scale 3D CBCT. Experiments on both simulated and real-world datasets show Riner outperforms existing SOTA supervised methods.

Mask Process Correction (MPC) is a critical step in advanced semiconductor manufacturing to mitigate pattern errors from e-beam writing and etching. However, conventional CPU-based commercial tools create a severe computational bottleneck as layout sizes and polygon complexity increase, resulting in prohibitively long MPC correction times. While academic efforts have explored using GPUs and advanced image-based machine learning algorithms on mask images, the inherent inefficiency of pixel mask representation results in an intractably large parameter space at the full-tile scale, limiting the practicality of these methods for industrial production. In this paper, we introduce a novel multimodal framework that processes point-cloud representations of GDS masks alongside E-beam Lithography (EBL) modeling information. This multi-modal approach enables massively parallel processing, preserves pattern fidelity, and ensures adaptability to variable e-beam lithography conditions. Experimental results show our method significantly reduces runtime while achieving edge placement error performance comparable to commercial tools.

Neural rendering enables photorealistic scene re-construction but remains difficult to deploy on edge devices due to intensive computation, redundant sampling, and memory bandwidth constraints. This work presents a high-performance neural rendering accelerator for real-time embedded rendering. The proposed design integrates: (1) a dual-process backend with fused micro-MLPs to significantly improve sample processing efficiency, (2) multi-resolution spatial partitioning with adaptive ray clustering to exploit sparsity and achieve over 95% cache hit rate, and (3) a multi-level scheduling framework with proactive prefetching to reduce MLP stalls. Implemented on FPGA, the prototype achieves 94.7 FPS at 800×800 resolution with 6.4 W power consumption. An ASIC implementation in 28 nm technology sustains 440 FPS at 268 mW. Experimental results demonstrate state-of-the-art performance and energy efficiency while preserving rendering quality above 30 dB PSNR.

3D Gaussian Splatting (3DGS) achieves photorealistic rendering but requires global sorting for α-blending, causing noticeable “popping” artifacts and hindering deployment on edge devices. While sort-free Order-Independent Transparency (OIT) methods circumvent sorting, they introduce “transparency” artifacts and suffer from inefficiency due to the absence of physical constraints. To address these limitations, we present Duplex-GS, a dual-hierarchy framework leveraging proxy-guided spatial organization and a novel hybrid renderer that combines α-blending with reformulated Weighted-Sum Rendering (WSR). We introduce explicit ellipsoidal cell proxies to encapsulate local Gaussians, which enables efficient proxy-level rasterization. This strategy drastically reduces the overhead associated with global sorting. Furthermore, we propose a physically grounded WSR scheme with cell-level early termination, which restores the physical constraints absent in prior OIT-based 3DGS methods, effectively eliminating both popping and transparency artifacts. Extensive experiments on diverse real-world benchmarks demonstrate the effectiveness of the OIT-based paradigm for 3DGS, enabled by a practical dual-hierarchy implementation. Quantitatively, our method delivers high-fidelity real-time rendering, outperforming prior OIT-based 3DGS methods by 1.5×– 4× in speed, while reducing radix-sort cost by 29.8%– 86.9% compared with conventional α-blending without compromising visual quality. Project page: https://duplexgs.github.io/.

Recent advances in generative models have democratized the creation of high-quality static 3D assets, yet animating these meshes remains a labor-intensive bottleneck. Traditional pipelines fracture this process into sequential stages—rigging, skinning, and motion synthesis—ignoring the inherent coupling between morphological structure and motor function. To bridge this gap, we introduce ACT, a unified generative framework that reformulates rigging and animation not as independent tasks, but as complementary views of a single hyper-kinematic process. Our key insight is to model the joint distribution of skeletal topology and temporal motion within a shared latent space. ACT utilizes a Vision Language Model (VLM) to extract semantic topological priors from arbitrary meshes, which then condition a Diffusion Transformer (DiT) backbone. By treating static rest poses and dynamic trajectories as a unified sequence, our model employs a task-aware masking strategy to flexibly perform zero-shot rigging, text-guided motion generation, and motion completion within a single end-to-end architecture. Furthermore, a geometry-guided decoder ensures that surface deformations are tightly coupled with the generated kinematics. Extensive experiments demonstrate that ACT generalizes robustly to diverse, non-humanoid characters without retraining. By replacing brittle cascaded pipelines with a holistic prior, our method enables novel applications such as semantic-driven topology editing and generative in-betweening, offering a versatile and efficient solution for automating 3D character animation.

Hair is a rich medium of visual and cultural expression, yet its digital modeling remains challenging due to the duality of fluidity and structure. Many existing generative approaches rely primarily on continuous diffusion fields, which entangle global topology with local texture and obscure the semantic and structural organization of hairstyles. To address this, we propose HairGPT, a strand-centric framework that treats strands as generative primitives and formulates realistic 3D hairstyle synthesis as a dual-decoupled autoregressive sequence modeling problem. Our method applies spatial decoupling across semantic scalp regions and structural decoupling along a hierarchical strand representation, progressing from global layout to fine-grained style. We further introduce a geometric tokenizer and region-aware semantic annotations to guide strand-level generation, enabling compositional editing, synthesis of rare and complex hairstyles, and adaptation to stylized domains. By aligning generative modeling with the workflow of digital grooming, HairGPT turns hair generation from opaque texture synthesis into a structured and semantically controllable authoring process, supporting robust semantic conditioning and high-fidelity results across realistic and stylized domains.

Capturing the full plenoptic light transport across spatial, angular, and temporal dimensions has long been a pursuit in computational imaging, yet it remains fundamentally constrained by the high dimensionality of the sampling space and the physical inaccessibility of scene regions due to self-occlusions. While time-resolved imaging records the temporal axis, existing methods are bottlenecked by the combinatorial complexity of the plenoptic function. This high dimensionality makes dense omni-dimensional sampling physically prohibitive. Simultaneously, tight coupling between illumination and viewpoint in current systems also precludes the full acquisition of plenoptic light transport. In this work, we present GenPIE, a Generative Plenoptic Imager designed to bridge the gap between sparse physical observations and high-dimensional light transport. We introduce a decoupled laser-detector hardware setup that enables independent control over illumination and detection, allowing for active probing of indirect light paths. To overcome the ill-posedness of sparse sampling and physical blind spots, we propose a generative inverse transient rendering framework. Our approach leverages 3D foundation models to provide strong semantic and 3D geometric priors for initialization, which are subsequently refined through a differentiable transient path tracer to ensure physically grounded adherence to the Transient Rendering Equation. We demonstrate that GenPIE supports a range of applications that are challenging for steady-state or purely neural methods, including disentangling multi-bounce light transport directly from captured transient videos, time unwarping, and time-resolved relighting. The project page is at https://wangzh1.github.io/GenPIE.

Recent advancements in autoregressive transformers have demonstrated remarkable potential for generating artist-quality meshes. However, the token ordering strategies employed by existing methods typically fail to meet professional artist standards, where coordinate-based sorting yields inefficiently long sequences, and patch-based heuristics disrupt the continuous edge flow and structural regularity essential for high-quality modeling. To address these limitations, we propose Strips as Tokens ( SATO ), a novel framework with a token ordering strategy inspired by triangle strips. By constructing the sequence as a connected chain of faces that explicitly encodes UV boundaries, our method naturally preserves the organized edge flow and semantic layout characteristic of artist-created meshes. A key advantage of this formulation is its unified representation, enabling the same token sequence to be decoded into either a triangle or quadrilateral mesh. This flexibility facilitates joint training on both data types: large-scale triangle data provides fundamental structural priors, while high-quality quad data enhances the geometric regularity of the outputs. Extensive experiments demonstrate that SATO consistently outperforms prior methods in terms of geometric quality, structural coherence, and UV segmentation.

Neural radiance fields (NeRFs) have transformed 3-D reconstruction and rendering, facilitating photorealistic image synthesis from sparse viewpoints. This work introduces an explicit data reuse neural rendering (EDR-NR) architecture, which reduces frequent external memory accesses (EMAs) and cache misses by exploiting the spatial locality from three phases, including rays, ray packets (RPs), and samples. The EDR-NR architecture features a four-stage scheduler that clusters rays on the basis of$Z$-order, prioritize lagging rays when ray divergence happens, reorders RPs based on spatial proximity, and issues samples out-of-orderly (OoO) according to the availability of on-chip feature data. In addition, a four-tier hierarchical RP marching (HRM) technique is integrated with an axis-aligned bounding box (AABB) to facilitate spatial skipping (SS), reducing redundant computations and improving throughput. Moreover, a balanced allocation strategy for feature storage is proposed to mitigate SRAM bank conflicts. Fabricated using a 40-nm process with a die area of 10.5 mm2, the EDR-NR chip demonstrates a$2.41\times $enhancement in normalized energy efficiency, a$1.21\times $improvement in normalized area efficiency, a$1.20\times $increase in normalized throughput, and a 53.42% reduction in on-chip SRAM consumption compared with state-of-the-art accelerators.

Sign languages, used by around 70 million Deaf individuals globally, are visual languages that convey visual and contextual information. Current methods in vision-based sign language recognition (SLR) and translation (SLT) struggle with dialogue scenes due to limited dataset diversity and the neglect of contextually relevant information. To address these challenges, we introduce SCOPE (Sign language COntextual Processing with Embedding from LLMs), a novel context-aware vision-based SLR and SLT framework. For SLR, we utilize dialogue contexts through a multi-modal encoder to enhance gloss-level recognition. For subsequent SLT, we further fine-tune a Large Language Model (LLM) by incorporating prior conversational context. We also contribute a new sign language dataset that contains 72 hours of Chinese sign language videos in contextual dialogues across various scenarios. Experimental results demonstrate that our SCOPE framework achieves state-of-the-art performance on multiple datasets, including Phoenix-2014T, CSL-Daily, and our SCOPE dataset. Moreover, surveys conducted with participants from the Deaf community further validate the robustness and effectiveness of our approach in real-world applications. Both our dataset and code will be open-sourced to facilitate further research.

We present Capturing the Unseen (CAPUS), a novel facial motion capture (MoCap) technique that operates without visual signals. CAPUS leverages miniaturized Inertial Measurement Units (IMUs) as a new sensing modality for facial motion capture. While IMUs have become essential in full-body MoCap for their portability and independence from environmental conditions, their application in facial MoCap remains underexplored. We address this by customizing micro-IMUs, small enough to be placed on the face, and strategically positioning them in alignment with key facial muscles to capture expression dynamics. CAPUS introduces the first facial IMU dataset, encompassing both IMU and visual signals from participants engaged in diverse activities such as multilingual speech, facial expressions, and emotionally intoned auditions. We train a Transformer Diffusion-based neural network to infer Blendshape parameters directly from IMU data. Our experimental results demonstrate that CAPUS reliably captures facial motion in conditions where visual-based methods struggle, including facial occlusions, rapid movements, and low-light environments. Additionally, by eliminating the need for visual inputs, CAPUS offers enhanced privacy protection, making it a robust solution for various applications.

The emergence of vision-language foundation models, such as CLIP, has revolutionized image-text representation, enabling a broad range of applications via prompt learning. Despite its promise, real-world datasets often contain noisy labels that can degrade prompt learning performance. In this paper, we demonstrate that using mean absolute error (MAE) loss in prompt learning, named PromptMAE, significantly enhances robustness against noisy labels while maintaining high accuracy. Though MAE is straightforward and recognized for its robustness, it is rarely used in noisy-label learning due to its slow convergence and poor performance outside prompt learning scenarios. To elucidate the robustness of PromptMAE, we leverage feature learning theory to show that MAE can suppress the influence of noisy samples, thereby improving the signal-to-noise ratio and enhancing overall robustness. Additionally, we introduce PromptOT, a prompt-based optimal transport data purification method to enhance the robustness further. PromptOT employs text features in vision-language models as prototypes to construct an optimal transportation matrix. This matrix effectively partitions datasets into clean and noisy subsets, allowing for the application of cross-entropy loss to the clean subset and MAE loss to the noisy subset. Our Noise-Label Prompt Learning method, named NLPrompt, offers a simple and efficient approach that leverages the expressive representations and precise alignment capabilities of vision-language models for robust prompt learning. We validate NLPrompt through extensive experiments across various noise settings, demonstrating significant performance improvements.

Diffusion models have demonstrated impressive capabilities in synthesizing diverse content. However, despite their high-quality outputs, these models often perpetuate social biases, including those related to gender and race. These biases can potentially contribute to harmful real-world consequences, reinforcing stereotypes and exacerbating inequalities in various social contexts. While existing research on diffusion bias mitigation has predominantly focused on guiding content generation, it often neglects the intrinsic mechanisms within diffusion models that causally drive biased outputs. In this paper, we investigate the internal processes of diffusion models, identifying specific decision-making mechanisms, termed bias features, embedded within the model architecture. By directly manipulating these features, our method precisely isolates and adjusts the elements responsible for bias generation, permitting granular control over the bias levels in the generated content. Through experiments on both unconditional and conditional diffusion models across various social bias attributes, we demonstrate our method’s efficacy in managing generation distribution while preserving image quality. We also dissect the discovered model mechanism, revealing different intrinsic features controlling fine-grained aspects of generation, boosting further research on mechanistic interpretability of diffusion models. The project website is at https://foundation-model-research.github.io/difflens.

Differentiable rendering enables efficient optimization by allowing gradients to be computed through the rendering process, facilitating 3D reconstruction, inverse rendering and neural scene representation learning. To ensure differentiability, existing solutions approximate or reformulate traditional rendering operations using smooth, probabilistic proxies such as volumes or Gaussian primitives. Consequently, they struggle to preserve sharp edges due to the lack of explicit boundary definitions. We present a novel hybrid representation, Bézier Gaussian Triangle (BG-Triangle), that combines Bézier triangle-based vector graphics primitives with Gaussian-Based probabilistic models, to maintain accurate shape modeling while conducting resolution-independent differentiable rendering. We present a robust and effective discontinuity-aware rendering technique to reduce uncertainties at object boundaries. We also employ an adaptive densification and pruning scheme for efficient training while reliably handling level-of-detail (LoD) variations. Experiments show that BG-Triangle achieves comparable rendering quality as 3DGS [27] but with superior boundary preservation. More importantly, BG-Triangle uses a much smaller number of primitives than its alternatives, showcasing the benefits of vectorized graphics primitives and the potential to bridge the gap between classic and emerging representations.

Diffusion-based policies have shown impressive performance in robotic manipulation tasks while struggling with out-of-domain distributions. Recent efforts attempted to enhance generalization by improving the visual feature encoding for diffusion policy. However, their generalization is typically limited to the same category with similar appearances. Our key insight is that leveraging affordances—manipulation priors that define "where" and "how" an agent interacts with an object—can substantially enhance generalization to entirely unseen object instances and categories. We introduce the Diffusion Policy with transferable Affordance (AffordDP), designed for generalizable manipulation across novel categories. AffordDP models affordances through 3D contact points and post-contact trajectories, capturing the essential static and dynamic information for complex tasks. The transferable affordance from in-domain data to unseen objects is achieved by estimating a 6D transformation matrix using foundational vision models and point cloud registration techniques. More importantly, we incorporate affordance guidance during diffusion sampling that can refine action sequence generation. This guidance directs the generated action to gradually move towards the desired manipulation for unseen objects while keeping the generated action within the manifold of action space. Experimental results from both simulated and real-world environments demonstrate that AffordDP consistently outperforms previous diffusion-based methods, successfully generalizing to unseen instances and categories where others fail.

3D affordance segmentation aims to link human instructions to touchable regions of 3D objects for embodied manipulations. Existing efforts typically adhere to single-object, single-affordance paradigms, where each affordance type or explicit instruction strictly corresponds to a specific affordance region and are unable to handle long-horizon tasks. Such a paradigm cannot actively reason about complex user intentions that often imply sequential affordances. In this paper, we introduce the Sequential 3D Affordance Reasoning task, which extends the traditional paradigm by reasoning from cumbersome user intentions and then decomposing them into a series of segmentation maps. Toward this, we construct the first instruction-based affordance segmentation benchmark that includes reasoning over both single and sequential affordances, comprising 180K instruction-point cloud pairs. Based on the benchmark, we propose our model, SeqAfford, to unlock the 3D multi-modal large language model with additional affordance segmentation abilities, which ensures reasoning with world knowledge and fine-grained affordance grounding in a cohesive framework. We further introduce a multi-granular language-point integration module to endow 3D dense prediction. Extensive experimental evaluations show that our model excels over well-established methods and exhibits open-world generalization with sequential reasoning abilities. Project page: https://seq-afford.github.io/.

A dexterous hand capable of grasping any object is essential for the development of general-purpose embodied intelligent robots. However, due to the high degree of freedom in dexterous hands and the vast diversity of objects, generating high-quality, usable grasping poses in a robust manner is a significant challenge. In this paper, we introduce DexGrasp Anything, a method that effectively integrates physical constraints into both the training and sampling phases of a diffusion-based generative model, achieving state-of-the-art performance across nearly all open datasets. Additionally, we present a new dexterous grasping dataset containing over 3.4 million diverse grasping poses for more than 15k different objects, demonstrating its potential to advance universal dexterous grasping. Code and dataset are available at https://github.com/4DVLab/DexGrasp-Anything

Moving toward the post-Moore era, full-chip mask optimization (MO) has become a pivotal step for semiconductor designers and manufacturers in extending current resolution enhancement techniques. The majority of recent research efforts have focused on clip-level restoration, employing a divide-and-conquer approach to mitigate the impacts of optical proximity and process bias across entire chips. Nevertheless, when confronted with industrial full-chip mask optimization challenges, these works exhibit limited correction capabilities, struggle with generalization, and are time-inefficient. In this paper, we propose a novel full-chip mask optimization paradigm based on a massive lithography data-driven large vision model. Our approach features a foundation layout feature extractor, which is aware of the mutual influence of polygons in long-range pattern perception as well as optical physics and chemical characteristics of lithography, matters. Compared with state-of-the-art (SOTA) works, our work demonstrates significant advantages in terms of resolution fidelity, correction speed, and the ability to handle full-chip scale layouts.

As integrated circuit (IC) feature sizes continue to shrink, using sub-resolution assist features (SRAF) becomes increasingly crucial for improving wafer pattern resolution and fidelity. However, model-based SRAF insertion techniques, while accurate, require substantial computational resources and are often impractical for industrial scenarios. This demands more efficient and industry-compatible methods that maintain high performance. In this work, we introduce LLM-SRAF, a novel framework for SRAF generation driven by a large language model fine-tuned on an SRAF dataset. LLM-SRAF accepts semantic prompt inputs, including SRAF generation task descriptions, OPC recipe, lithography conditions, mask rules, and sequential layout descriptions, to directly generate SRAFs. Both supervised fine-tuning and reinforcement learning with human feedback (RLHF) are employed to enable the model to acquire domain-specific knowledge and specialize in SRAF generation. Experimental results show that LLM-SRAF outperforms existing state-of-the-art methods in metrics of mask quality, including edge placement error (EPE) and process variation band (PVB) area. Moreover, the runtime of LLM-SRAF is also 3x faster compared to the Calibre commercial tool.

The increasing complexity of modern SoC designs creates challenges in efficiently exploring vast design spaces. Current approaches often reduce microarchitectures to simple parameter vectors, overlooking their rich information embedded in both functional behaviors and implementation details. This paper proposes an LLM-augmented multi-modal fusion method that captures this dual nature of microarchitecture design. By recombining design parameters with their natural language descriptors, we leverage a domain-knowledge-enhanced LLM to extract semantic features that represent functional behavior. Simultaneously, we process Chisel-compiled RTL through a graph neural network to capture structural implementation details. This multi-modal approach enables more effective feature extraction from limited evaluation data. We integrate these rich features into an MLP enhanced with Monte Carlo dropout. This approach provides uncertainty quantification while enabling end-to-end training, allowing the pre-trained feature extractors to be fine-tuned during exploration through Bayesian optimization. Experimental results and ablation studies on a Gemmini-based RISC-V SoC demonstrate that our approach significantly improves exploration efficiency and prediction quality under data limitations.

As IC fabrication advances toward smaller process nodes, design technology co-optimization (DTCO) has emerged as a critical enabler of chip performance advancements. Lithography simulation, vital for bridging design and manufacturing, now plays an indispensable role in designing litho-friendly layouts/masks and developing resolution enhancement techniques (RETs). While academia and industry have explored statistical techniques and machine learning models for simulators, the computing paradigm and hardware prevent these solutions from efficiently and accurately simulating the complicated optical imaging coupled with resist film imaging. In this paper, we propose a new simulation paradigm: LMLitho (large vision model-driven lithography simulator), trained on circa one hundred thousand triplets of illumination maps, masks, and resist images. The cross-attention mechanism in our simulator inherently captures diffraction patterns akin to light wave interference within mask features, while hierarchical attention layers enable the modeling of long-range diffraction effects (e.g., proximity effects). A comprehensive dataset encompassing diverse classical types of source and mask patterns, including both metal-1 and via layers, is generated to meet the requirements of training our large vision model-based simulator1. The experimental results demonstrate that our simulator achieves over 120× speedup compared to existing commercial solutions while preserving comparable high fidelity, and exhibits superior generalization to advanced process nodes. When deployed in inverse lithography technology (ILT)-guided mask optimization workflows, masks of higher quality are generated than existing solutions.

Critical pattern selection is an essential foundation for lithography technologies such as full-chip Source Mask Optimization (SMO) and Optical Proximity Correction (OPC) model calibration. Traditional methods are mostly based on image feature extraction and clustering, without integrating lithography domain knowledge, and they suffer from high time complexity and excessive computational resource consumption. As a result, they fail to fully capture the optical characteristics of complex patterns and lack subsequent verification of the "critical" nature of the patterns. Moreover, past work often relied on manually annotated data, and the generalization ability needs further validation. To overcome these limitations, this paper proposes a semi-supervised critical pattern selection method based on frequency domain and LVM collaboration. This method defines the critical patterns by describing the optical characteristics of the patterns in the frequency domain and uses the ViT model to extract the nonlinear features of the patterns, ultimately selecting the critical patterns efficiently. This approach meets the demands of high-precision and high-efficiency lithography techniques for advanced processes. A series of simulations and comparisons performed using commercial tool, verify the validity of the proposed method.

Real-time synthesis of physically plausible human interactions remains a critical challenge for immersive VR/AR systems and humanoid robotics. While existing methods demonstrate progress in kinematic motion generation, they often fail to address the fundamental tension between real-time responsiveness, physical feasibility, and safety requirements in dynamic human-machine interactions. We introduce Human-X, a novel framework designed to enable immersive and physically plausible human interactions across diverse entities, including human-avatar, human-humanoid, and human-robot systems. Unlike existing approaches that focus on post-hoc alignment or simplified physics, our method jointly predicts actions and reactions in real-time using an auto-regressive reaction diffusion planner, ensuring seamless synchronization and context-aware responses. To enhance physical realism and safety, we integrate an actor-aware motion tracking policy trained with reinforcement learning, which dynamically adapts to interaction partners' movements while avoiding artifacts like foot sliding and penetration. Extensive experiments on the Inter-X and InterHuman datasets demonstrate significant improvements in motion quality, interaction continuity, and physical plausibility over state-of-the-art methods. Our framework is validated in real-world applications, including virtual reality interface for human-robot interaction, showcasing its potential for advancing human-robot collaboration.

High quality and high speed videography using Non-Line-of-Sight (NLOS) imaging benefit autonomous navigation, collision prevention, and post-disaster search and rescue tasks. Current solutions have to balance between the frame rate and image quality. High frame rates, for example, can be achieved by reducing either per-point scanning time or scanning density, but at the cost of lowering the information density at individual frames. Fast scanning process further reduces the signal-to-noise ratio and different scanning systems exhibit different distortion characteristics. In this work, we design and employ a new Transient Transformer architecture called TransiT to achieve real-time NLOS recovery under fast scans. TransiT directly compresses the temporal dimension of input transients to extract features, reducing computation costs and meeting high frame rate requirements. It further adopts a feature fusion mechanism as well as employs a spatial-temporal Transformer to help capture features of NLOS transient videos. Moreover, TransiT applies transfer learning to bridge the gap between synthetic and real-measured data. In real experiments, TransiT manages to reconstruct from sparse transients of $16 \times 16$ measured at an exposure time of 0.4 ms per point to NLOS videos at a $64 \times 64$ resolution at 10 frames per second. We will make our code and dataset available to the community.

Soccer is a globally renowned sport with significant applications in video games and VR/AR. However, generating realistic soccer motions remains challenging due to the intricate interactions between the human player and the ball. In this paper, we introduce SMGDiff, a novel two-stage framework for generating real-time and user-controllable soccer motions. Our key idea is to integrate real-time character control with a powerful diffusion-based generative model, ensuring high-quality and diverse output motion. In the first stage, we instantly transform coarse user controls into diverse global trajectories of the character. In the second stage, we employ a transformer-based autoregressive diffusion model to generate soccer motions based on trajectory conditioning. We further incorporate a contact guidance module during inference to optimize the contact details for realistic ball-foot interactions. Moreover, we contribute a large-scale soccer motion dataset consisting of over 1.08 million frames of diverse soccer motions. Extensive experiments demonstrate that our SMGDiff significantly outperforms existing methods in terms of motion quality and condition alignment.

Pose estimation from unordered images is fundamental for 3D reconstruction, robotics, and scientific imaging. Recent geometric foundation models, such as DUSt3R, enable end-to-end dense 3D reconstruction but remain underexplored in scientific imaging fields like cryo-electron microscopy (cryo-EM) for near-atomic protein reconstruction. In cryo-EM, pose estimation and 3D reconstruction from unordered particle images still depend on time-consuming iterative optimization, primarily due to challenges such as low signal-to-noise ratios (SNR) and distortions from the contrast transfer function (CTF). We introduce CryoFastAR, the first geometric foundation model that can directly predict poses from Cryo-EM noisy images for Fast ab initio Reconstruction. By integrating multi-view features and training on large-scale simulated cryo-EM data with realistic noise and CTF modulations, CryoFastAR enhances pose estimation accuracy and generalization. To enhance training stability, we propose a progressive training strategy that first allows the model to extract essential features under simpler conditions before gradually increasing difficulty to improve robustness. Experiments show that CryoFastAR achieves comparable quality while significantly accelerating inference over traditional iterative approaches on both synthetic and real datasets.

Vision Transformer models exhibit immense power yet remain opaque to human understanding, posing challenges and risks for practical applications. While prior research has attempted to demystify these models through input attribution and neuron role analysis, there's been a notable gap in considering layer-level information and the holistic path of information flow across layers. In this paper, we investigate the significance of influential neuron paths within vision Transformers, which is a path of neurons from the model input to output that impacts the model inference most significantly. We first propose a joint influence measure to assess the contribution of a set of neurons to the model outcome. And we further provide a layer-progressive neuron locating approach that efficiently selects the most influential neuron at each layer trying to discover the crucial neuron path from input to output within the target model. Our experiments demonstrate the superiority of our method finding the most influential neuron path along which the information flows, over the existing baseline solutions. Additionally, the neuron paths have illustrated that vision Transformers exhibit some specific inner working mechanism for processing the visual information within the same image category. We further analyze the key effects of these neurons on the image classification task, showcasing that the found neuron paths have already preserved the model capability on downstream tasks, which may also shed some lights on real-world applications like model pruning. The project website including implementation code is available at https://foundation-model-research.github.io/NeuronPath/.

Motion correction (MoCo) in radial MRI is a particularly challenging problem due to the unpredictability of subject movement. Current state-of-the-art (SOTA) MoCo algorithms often rely on extensive high-quality MR images to pre-train neural networks, which constrains the solution space and leads to outstanding image reconstruction results. However, the need for large-scale datasets significantly increases costs and limits model generalization. In this work, we propose Moner, an unsupervised MoCo method that jointly reconstructs artifact-free MR images and estimates accurate motion from undersampled, rigid motion-corrupted k-space data, without requiring any training data. Our core idea is to leverage the continuous prior of implicit neural representation (INR) to constrain this ill-posed inverse problem, facilitating optimal solutions. Specifically, we integrate a quasi-static motion model into the INR, granting its ability to correct subject's motion. To stabilize model optimization, we reformulate radial MRI reconstruction as a back-projection problem using the Fourier-slice theorem. Additionally, we propose a novel coarse-to-fine hash encoding strategy, significantly enhancing MoCo accuracy. Experiments on multiple MRI datasets show our Moner achieves performance comparable to SOTA MoCo techniques on in-domain data, while demonstrating significant improvements on out-of-domain data. The code is available at: https://github.com/iwuqing/Moner

This paper addresses the challenging task of generating dynamic Human-Object Interactions from textual descriptions, named Text2HOI. While most existing works assume interactions with limited body parts or static objects, our task involves addressing the variation in human motion, the diversity of object shapes, and the semantic vagueness of object motion simultaneously. To tackle this, we propose a novel Text-guided Human-Object Interaction diffusion model with Relation Intervention (THOR). THOR is a cohesive diffusion model equipped with a relation intervention mechanism. In each diffusion step, we initiate text-guided human and object motion and then leverage human-object relations to intervene in object motion. This intervention enhances the spatial-temporal relations between humans and objects, with human-centric motion providing additional guidance for synthesizing consistent motion from text. To achieve more reasonable and realistic results, relation intervention loss is introduced at different levels of motion granularity.

Diffusion models have emerged as mainstream framework in visual generation. Building upon this success, the integration of Mixture of Experts (MoE) methods has shown promise in enhancing model scalability and performance. In this paper, we introduce Race-DiT, a novel MoE model for diffusion transformers with a flexible routing strategy, Expert Race. By allowing tokens and experts to compete together and select the top candidates, the model learns to dynamically assign experts to critical tokens. Additionally, we propose per-layer regularization to address challenges in shallow layer learning, and router similarity loss to prevent mode collapse, ensuring better expert utilization. Extensive experiments on ImageNet validate the effectiveness of our approach, showcasing significant performance gains while promising scaling properties.

Recent advances in diffusion bridge models leverage Doob’s $h$-transform to establish fixed endpoints between distributions, demonstrating promising results in image translation and restoration tasks. However, these approaches frequently produce blurred or excessively smoothed image details and lack a comprehensive theoretical foundation to explain these shortcomings. To address these limitations, we propose UniDB, a unified framework for diffusion bridges based on Stochastic Optimal Control (SOC). UniDB formulates the problem through an SOC-based optimization and derives a closed-form solution for the optimal controller, thereby unifying and generalizing existing diffusion bridge models. We demonstrate that existing diffusion bridges employing Doob’s $h$-transform constitute a special case of our framework, emerging when the terminal penalty coefficient in the SOC cost function tends to infinity. By incorporating a tunable terminal penalty coefficient, UniDB achieves an optimal balance between control costs and terminal penalties, substantially improving detail preservation and output quality. Notably, UniDB seamlessly integrates with existing diffusion bridge models, requiring only minimal code modifications. Extensive experiments across diverse image restoration tasks validate the superiority and adaptability of the proposed framework. Our code is available at https://github.com/UniDB-SOC/UniDB/.

This paper presents a novel Expressive Facial Control (ExFace) method based on Diffusion Transformers, which achieves precise mapping from human facial blendshapes to bionic robot motor control. By incorporating an innovative model bootstrap training strategy, our approach not only generates high-quality facial expressions but also significantly improves accuracy and smoothness. Experimental results demonstrate that the proposed method outperforms previous methods in terms of accuracy, frames per second (FPS), and response time. Furthermore, we develop the ExFace dataset driven by human facial data. ExFace shows excellent real-time performance and natural expression rendering in applications such as robot performances and human-robot interactions, offering a new solution for bionic robot interaction.

Lithography orchestrates a symphony of light, mask and photochemicals to transfer the integrated circuit patterns onto the wafer. Lithography simulation serves as the critical nexus between circuit design and manufacturing, where its speed and accuracy fundamentally govern the optimization quality of downstream resolution enhancement techniques (RET). While machine learning promises to circumvent computational limitations of lithography process through data-driven or physics-informed approximations of computational lithography, existing simulators suffer from inadequate lithographic awareness due to insufficient training data capturing essential process variations and mask correction rules. We present LithoSim, the most comprehensive lithography simulation benchmark to date, featuring over $4$ million high-resolution input-output pairs with rigorous physical correspondence. The dataset systematically incorporates alterable optical source distributions, metal and via mask topologies with optical proximity correction (OPC) variants, and process windows reflecting fab-realistic variations. By integrating domain-specific metrics spanning AI performance and lithographic fidelity, LithoSim establishes a unified evaluation framework for data-driven and physics-informed computational lithography. The data (https://huggingface.co/datasets/grandiflorum/LithoSim), code (https://dw-hongquan.github.io/LithoSim), and pre-trained models (https://huggingface.co/grandiflorum/LithoSim) are released openly to support the development of hybrid ML-based and high-fidelity lithography simulation for the benefit of semiconductor manufacturing.

Understanding objects at the level of their constituent parts is fundamental to advancing computer vision, graphics, and robotics. While datasets like PartNet have driven progress in 3D part understanding, their reliance on untextured geometries and expert-dependent annotation limits scalability and usability. We introduce PartNeXt, a next-generation dataset addressing these gaps with over 23000 high-quality, textured 3D models annotated with fine-grained, hierarchical part labels across 50 categories. We benchmark PartNeXt on two tasks: (1) class-agnostic part segmentation, where state-of-the-art methods (e.g., PartField, SAMPart3D) struggle with fine-grained and leaf-level parts, and (2) 3D part-centric question answering, a new benchmark for 3D-LLMs that reveals significant gaps in open-vocabulary part grounding. Additionally, training Point-SAM on PartNeXt yields substantial gains over PartNet, underscoring the dataset’s superior quality and diversity. By combining scalable annotation, texture-aware labels, and multi-task evaluation, PartNeXt opens new avenues for research in structured 3D understanding.

Manhattan representations, defined by axis-aligned, orthogonal structures, are widely used in vision, robotics, and semiconductor design for their geometric regularity and algorithmic simplicity. In integrated circuit (IC) design, Manhattan geometry is key for routing, design rule checking, and lithographic manufacturability. However, as feature sizes shrink, optical system distortions lead to inconsistency between intended layout and printed wafer. Although Inverse Lithography Technology(ILT) is proposed to compensates these effects, learning-based ILT methods, while achieving high simulation fidelity, often generate curvilinear masks on continuous pixel grids, violating Manhattan constraints. Therefore, we propose TokMan, the first framework to formulate mask optimization as a discrete, structure-aware sequence modeling task. Our method leverages a Diffusion Transformer to tokenize layouts into discrete geometric primitives with polygon-wise dependencies and denoise Manhattan-aligned point sequences corrupted by optical proximity effects, while ensuring binary, manufacturable masks. Trained with self-supervised lithographic feedback through differentiable simulation and refined with ILT post-processing, TokMan achieves state-of-the-art fidelity, runtime efficiency, and strict manufacturing compliance on a large-scale dataset of IC layouts.

Accurate and efficient modeling of large-scale urban scenes is critical for applications such as AR navigation, UAV-based inspection, and smart city digital twins. While aerial imagery offers broad coverage and complements limitations of ground-based data, reconstructing city-scale environments from such views remains challenging due to occlusions, incomplete geometry, and high memory demands. Recent advances like 3D Gaussian Splatting (3DGS) improve scalability and visual quality but remain limited by dense primitive usage, long training times, and poor suitability for edge devices. We propose CityGo, a hybrid framework that combines textured proxy geometry with residual and surrounding 3D Gaussians for lightweight, photorealistic rendering of urban scenes from aerial perspectives. Our approach first extracts compact building proxy meshes from MVS point clouds, then uses zero-order SH Gaussians to generate occlusion-free textures via image-based rendering and back-projection. To capture high-frequency details, we introduce residual Gaussians placed based on proxy-photo discrepancies and guided by depth priors. Broader urban context is represented by surrounding Gaussians, with importance-aware downsampling applied to non-critical regions to reduce redundancy. A tailored optimization strategy jointly refines proxy textures and Gaussian parameters, enabling real-time rendering of complex urban scenes on mobile GPUs with significantly reduced training and memory requirements. Extensive experiments on real-world aerial datasets demonstrate that our hybrid representation achieves fastest training speed, while delivering comparable visual fidelity to pure 3D Gaussian Splatting approaches. Furthermore, CityGo enables real-time rendering of large-scale urban scenes on mobile consumer GPUs, with substantially reduced memory usage and energy consumption.

We introduce HOLI-1-to-3, a novel technique for holistic 3D shape recovery from a single-viewpoint input, by effectively combining line-of-sight (LOS) and non-line-of-sight (NLOS) imaging. We leverage advancements in ultrafast time-of-flight (ToF) sensors and learning-based 3D shape inference techniques, such as diffusion models. HOLI-1-to-3 employs a new neural plenoptic representation, which unifies radiance fields (for LOS RGB images) and transient fields (for NLOS transients). HOLI-1-to-3 is optimized through a two-stage pipeline involving diffusion priors and transients prior. Our technique allows for accurate and continuous reconstruction of both visible and invisible parts of objects from a single view. Comprehensive experiments on both simulated and real-world datasets demonstrate the effectiveness of HOLI-1-to-3in resolving ambiguities in invisible parts of objects and significantly improving overall generation quality. The datasets used in our experiments will be made available to the research community to facilitate further achievements in holistic 3D shape recovery.

Recovering high-quality 3D scenes from a single RGB image is a challenging task in computer graphics. Current methods often struggle with domain-specific limitations or low-quality object generation. To address these, we propose CAST (Component-Aligned 3D Scene Reconstruction from a Single RGB Image), a novel method for 3D scene reconstruction. CAST starts by extracting object-level 2D segmentation and relative depth information from the input image, followed by using a GPT-based model to analyze inter-object spatial relations. This enables understanding of how objects relate to each other within the scene, ensuring more coherent reconstruction. CAST then employs an occlusion-aware large-scale 3D generation model to independently generate each object's full geometry, using Masked Auto Encoder (MAE) and point cloud conditioning to mitigate the effects of occlusions and partial object information, ensuring accurate alignment with the source image's geometry and texture. To align each object with the scene, the alignment generation model computes the necessary transformations, allowing the generated meshes to be accurately placed and integrated into the scene's point cloud. Finally, CAST applies a physics-aware correction mechanism, which leverages a fine-grained relation graph to generate a constraint graph. This graph guides the optimization of object poses, ensuring physical consistency and spatial coherence. By utilizing Signed Distance Fields (SDF), the model effectively addresses issues such as occlusions, object penetration, and floating objects, ensuring that the generated scene accurately reflects real-world physical interactions. Experimental results demonstrate that CAST significantly improves the quality of single-image 3D scene reconstruction, offering enhanced realism and accuracy in scene understanding and reconstruction tasks. CAST has practical applications in virtual content creation, such as immersive game environments and film production, where real-world setups can be seamlessly integrated into virtual landscapes. Additionally, CAST can be leveraged in robotics, enabling efficient real-to-simulation workflows and providing realistic, scalable simulation environments for robotic systems.

3D creation has always been a unique human strength, driven by our ability to deconstruct and reassemble objects using our eyes, mind and hand. However, current 3D design tools struggle to replicate this natural process, requiring considerable artistic expertise and manual labor. This paper introduces BANG, a novel generative approach that bridges 3D generation and reasoning, allowing for intuitive and flexible part-level decomposition of 3D objects. At the heart of BANG is "Generative Exploded Dynamics", which creates a smooth sequence of exploded states for an input geometry, progressively separating parts while preserving their geometric and semantic coherence. BANG utilizes a pre-trained large-scale latent diffusion model, fine-tuned for exploded dynamics with a lightweight exploded view adapter, allowing precise control over the decomposition process. It also incorporates a temporal attention module to ensure smooth transitions and consistency across time. BANG enhances control with spatial prompts, such as bounding boxes and surface regions, enabling users to specify which parts to decompose and how. This interaction can be extended with multimodal models like GPT-4, enabling 2D-to-3D manipulations for more intuitive and creative workflows. The capabilities of BANG extend to generating detailed part-level geometry, associating parts with functional descriptions, and facilitating component-aware 3D creation and manufacturing workflows. Additionally, BANG offers applications in 3D printing, where separable parts are generated for easy printing and reassembly. In essence, BANG enables seamless transformation from imaginative concepts to detailed 3D assets, offering a new perspective on creation that resonates with human intuition.

This paper introduces the concept of Content-Aware Radiance Fields (CoARF), which adaptively aligns the model complexity with the scene intricacy. By examining the intricacies of radiance fields from three perspectives, model complexity is adapted through scalable feature grids, dynamic neural networks, and model quantization. Specifically, we propose a hash collision detection mechanism that removes redundant feature grid by restricting the valid hash collision to reasonable level, making the space complexity scalable. We introduce an uncertainty-aware decoded layer, where simple points are early-exited to prevent them from being processed by deeper network layers, ensuring computational complexity scalable. Furthermore, we propose Learned Bitwidth Quantization (LBQ) and Adversarial Content-Aware Quantization (A-CAQ) paradigms by making the bitwidth of parameters differentiable and trainable, allowing for adjustable quantization schemes. Building on these techniques, the proposed CoARF++ framework enables a scalable pipeline for radiance fields that is tailored to the unique characteristics of scene complexity and quality requirement. Extensive experiments demonstrate a significant and adjustable reduction in model complexity across various NeRF variants, while maintaining the necessary reconstruction and rendering quality, making it advantageous for the practical deployment of radiance field models.

Neural rendering, a transformative approach for 3-D scene reconstruction and rendering, has advanced rapidly in recent years. This article introduces an energy-efficient neural rendering coprocessor that implements the popular and widely used instant neural graphics primitive (Instant-NGP) algorithm. In particular, we address the challenges of limited resources for deploying Instant-NGP on edge by proposing a dedicated architecture, which incorporates three main innovations: 1) we optimize occupancy grid queries in the ray marching module by partitioning the grid and decoupling the query process from sampling point generation, which improves both efficiency and memory usage; 2) we introduce a bilinked list-based ray switching strategy, which ensures continuous pipeline utilization to overcome the inefficiencies caused by sequential processing; and 3) we optimize the hash encoding process by incorporating quantization-aware training (QAT), enabling the hash table to fit into on-chip memory, thereby improving performance on resource-constrained devices. To demonstrate the effectiveness of our architecture, we design and fabricate a proof-of-concept chip using 40-nm CMOS technology and develop a testing system to evaluate its performance. Measurement results validate the advantages of the proposed design, showing that our chip achieves superior energy efficiency compared to both server and edge graphics processing units (GPUs), as well as other state-of-the-art neural rendering chip designs.

We have recently seen tremendous progress in photo-real human modeling and rendering. Yet, efficiently ren-dering realistic human performance and integrating it into the rasterization pipeline remains challenging. In this pa-per, we present HiFi4G, an explicit and compact Gaussian-based approach for high-fidelity human performance ren-dering from dense footage. Our core intuition is to marry the 3D Gaussian representation with non-rigid tracking, achieving a compact and compression-friendly representation. We first propose a dual-graph mechanism to obtain motion priors, with a coarse deformation graph for effective initialization and a fine-grained Gaussian graph to en-force subsequent constraints. Then, we utilize a 4D Gaus-sian optimization scheme with adaptive spatial-temporal regularizers to effectively balance the non-rigid prior and Gaussian updating. We also present a companion compression scheme with residual compensation for immersive experiences on various platforms. It achieves a substantial compression rate of approximately 25 times, with less than 2MB of storage per frame. Extensive experiments demon-strate the effectiveness of our approach, which significantly outperforms existing approaches in terms of optimization speed, rendering quality, and storage overhead. Project page: https://nowheretrix.github.io/HiFi4G/.

We have recently seen tremendous progress in realistic text-to-motion generation. Yet, the existing methods of-ten fail or produce implausible motions with unseen text inputs, which limits the applications. In this paper, we present OMG, a novel framework, which enables compelling motion generation from zero-shot open-vocabulary text prompts. Our key idea is to carefully tailor the pretrain-then-finetune paradigm into the text-to-motion generation. At the pre-training stage, our model improves the gener-ation ability by learning the rich out-of-domain inherent motion traits. To this end, we scale up a large unconditional diffusion model up to 1B parameters, so as to utilize the massive unlabeled motion data up to over 20M motion instances. At the subsequent fine-tuning stage, we intro-duce motion ControlNet, which incorporates text prompts as conditioning information, through a trainable copy of the pre-trained model and the proposed novel Mixture-of-Controllers (MoC) block. MoC block adaptively rec-ognizes various ranges of the sub-motions with a cross-attention mechanism and processes them separately with the text-token-specific experts. Such a design effectively aligns the CLIP token embeddings of text prompts to var-ious ranges of compact and expressive motion features. Ex-tensive experiments demonstrate that our OMG achieves significant improvements over the state-of-the-art meth-ods on zero-shot text-to-motion generation. Project page: https://tr3e.github.io/omg-page.

For human-centric large-scale scenes, fine-grained modeling for 3D human global pose and shape is significant for scene understanding and can benefit many real-world applications. In this paper, we present LiveHPS, a novel single-LiDAR-based approach for scene-level Human Pose and Shape estimation without any limitation of light conditions and wearable devices. In particular, we design a distillation mechanism to mitigate the distribution-varying effect of LiDAR point clouds and exploit the temporal-spatial geometric and dynamic information existing in consecutive frames to solve the occlusion and noise disturbance. LiveHPS, with its efficient configuration and high-quality output, is well-suited for real-world applications. More-over, we propose a huge human motion dataset, named FreeMotion, which is collected in various scenarios with diverse human poses, shapes and translations. It consists of multimodal and multi-view acquisition data from calibrated and synchronized LiDARs, cameras, and IMUs. Extensive experiments on our new dataset and other public datasets demonstrate the SOTA performance and robustness of our approach.

Estimating full-body human motion via sparse tracking signals from head-mounted displays and hand controllers in 3D scenes is crucial to applications in AR/VR. One of the biggest challenges to this task is the one-to-many mapping from sparse observations to dense full-body motions, which endowed inherent ambiguities. To help resolve this ambiguous problem, we introduce a new framework to combine rich contextual information provided by scenes to benefit fullbody motion tracking from sparse observations. To estimate plausible human motions given sparse tracking signals and 3D scenes, we develop S2Fusion, a unified framework fusing Scene and sparse Signals with a conditional difFusion model. S2Fusion first extracts the spatial-temporal relations residing in the sparse signals via a periodic autoencoder, and then produces time-alignment feature embedding as additional inputs. Subsequently, by drawing initial noisy motion from a pre-trained prior, S2Fusion utilizes conditional diffusion to fuse scene geometry and sparse tracking signals to generate full-body scene-aware motions. The sampling procedure of S2Fusion is further guided by a specially designed scene-penetration loss and phase-matching loss, which effectively regularizes the motion of the lower body even in the absence of any tracking signals, making the generated motion much more plausible and coherent. Extensive experimental results have demonstrated that our S2Fusion outperforms the state-of-the-art in terms of estimation quality and smoothness. Code is available at https://github.com/jntang/S2Fusion.

Neural Radiance Fields (NeRFs) excel in photorealistically rendering static scenes. However, rendering dynamic, long-duration radiance fields on ubiquitous devices remains challenging, due to data storage and computational constraints. In this paper, we introduce VideoRF, the first approach to enable real-time streaming and rendering of dynamic human-centric radiance fields on mobile platforms. At the core is a serialized 2D feature image stream representing the 4D radiance field all in one. We introduce a tailored training scheme directly applied to this 2D domain to impose the temporal and spatial redundancy of the feature image stream. By leveraging the redundancy, we show that the feature image stream can be efficiently compressed by 2D video codecs, which allows us to exploit video hardware accelerators to achieve real-time decoding. On the other hand, based on the feature image stream, we propose a novel rendering pipeline for VideoRF, which has specialized space mappings to query radiance properties efficiently. Paired with a deferred shading model, VideoRF has the capability of real-time rendering on mobile devices thanks to its efficiency. We have developed a real-time interactive player that enables online streaming and rendering of dynamic scenes, offering a seamless and immersive free-viewpoint experience across a range of devices, from desktops to mobile phones. Our project page is available at https://aoliao12138.github.io/VideoRF/.

The recently emerging text-to-motion advances have inspired numerous attempts for convenient and interactive human motion generation. Yet, existing methods are largely limited to generating body motions only without considering the rich two-hand motions, let alone handling various conditions like body dynamics or texts. To break the data bottleneck, we propose BOTH57M, a novel multi-modal dataset for two-hand motion generation. Our dataset includes accurate motion tracking for the human body and hands and provides pair-wised finger-level hand annotations and body descriptions. We further provide a strong baseline method, BOTH2Hands, for the novel task: generating vivid two-hand motions from both implicit body dynamics and explicit text prompts. We first warm up two parallel body-to-hand and text-to-hand diffusion models and then utilize the cross-attention transformer for motion blending. Extensive experiments and cross-validations demonstrate the effectiveness of our approach and dataset for generating convincing two-hand motions from the hybrid body-and-textual conditions. Our dataset and code will be released to the community for future research, which can be found at github.

Humans naturally interact with both others and the sur-rounding multiple objects, engaging in various social activi-ties. However, recent advances in modeling human-object interactions mostly focus on perceiving isolated individuals and objects, due to fundamental data scarcity. In this pa-per, we introduce HOI-M3, a novel large-scale dataset for modeling the interactions of Multiple huMans and Multiple objects. Notably, it provides accurate 3D tracking for both humans and objects from dense RGB and object-mounted IMU inputs, covering 199 sequences and 181M frames of diverse humans and objects under rich activities. With the unique HOI-M3dataset, we introduce two novel data-driven tasks with companion strong baselines: monocular capture and unstructured generation of multiple human-object interactions. Extensive experiments demonstrate that our dataset is challenging and worthy of further research about multiple human-object interactions and behavior analysis. Our HOI-M3 dataset, corresponding codes, and pre-trained models will be disseminated to the community for future research, which can be found at https://juzezhang.github.io/HOIM3_ProjectPage/

We are living in a world surrounded by diverse and “smart” devices with rich modalities of sensing ability. Conveniently capturing the interactions between us humans and these objects remains far-reaching. In this paper, we present I'm-HOI, a monocular scheme to faithfully capture the 3D motions of both the human and object in a novel setting: using a minimal amount of RGB camera and object-mounted Inertial Measurement Unit (IMU). It combines general motion inference and category-aware refinement. For the former, we introduce a holistic human-object tracking method to fuse the IMU signals and the RGB stream and progressively recover the human motions and subsequently the companion object motions. For the latter, we tailor a category-aware motion diffusion model, which is conditioned on both the raw IMU observations and the results from the previous stage under over-parameterization representation. It significantly refines the initial results and generates vivid body, hand, and object motions. Moreover, we contribute a large dataset with ground truth human and object motions, dense RGB inputs, and rich object-mounted IMU measurements. Exten-sive experiments demonstrate the effectiveness of I'm-HOI under a hybrid capture setting. Our dataset and code will be released to the community.

Layout hotspot detection approaches are challenged by the time-to-market constraint and complex designs under rapid downscaling of technology nodes. Pattern matching and learning-based detectors are proposed as quick detection methods. These layout image-based detectors use images transformed from binary database files of layout like GDSII as their inputs. Italy leads to foreground information (e.g., metal polygons) loss and even distortion when shrinking the image size to fit the approach input. Moreover, plenty of irrelevant background information such as non-polygon pixels is also fed into the model, which hinders the fitting of the model and results in a waste of computational resources. In this work, for the first time, we propose a new layout hotspot detection paradigm, where hotspots are directly detected on binary database files by exploiting a hierarchical GDS semantic representation scheme and a well-designed pre-trained natural language processing (NLP) model. Compared with state-of-the-art (SOTA) works, the proposed detector achieves better results on both the ICCAD2012 metal layer benchmark and the more challenging ICCAD2020 via layer benchmark, which demonstrates the effectiveness and efficiency.

Neural Volume Rendering (NVR), a novel paradigm for the longstanding problem of photo-realistic rendering of virtual worlds, has developed explosively in the past three years. The unique and substantial computational requirements of NVR pose challenge on deploying NVR to existing dedicated accelerator for neural networks. In this work, we propose ZeroTetris, a spacial feature similarity-based sparse multilayer perceptron (MLP) hardware accelerator for NVR. By leveraging the unique similarity-based sparsity between adjacent sampling points in NVR models, ZeroTetris efficiently bypass the computation of zero activations, thereby enhancing energy efficiency. Evaluation results affirm the effectiveness of the proposed design, showcasing ZeroTetris's superior performance in both area and power efficiency compared to other dedicated sparse matrix multiplication or MLP accelerator designs.

Recent advances in diffusion models attempt to handle conditional generative tasks by utilizing a differentiable loss function for guidance without the need for additional training. While these methods achieved certain success, they often compromise on sample quality and require small guidance step sizes, leading to longer sampling processes. This paper reveals that the fundamental issue lies in the manifold deviation during the sampling process when loss guidance is employed. We theoretically show the existence of manifold deviation by establishing a certain lower bound for the estimation error of the loss guidance. To mitigate this problem, we propose Diffusion with Spherical Gaussian constraint (DSG), drawing inspiration from the concentration phenomenon in high-dimensional Gaussian distributions. DSG effectively constrains the guidance step within the intermediate data manifold through optimization and enables the use of larger guidance steps. Furthermore, we present a closed-form solution for DSG denoising with the Spherical Gaussian constraint. Notably, DSG can seamlessly integrate as a plugin module within existing training-free conditional diffusion methods. Implementing DSG merely involves a few lines of additional code with almost no extra computational overhead, yet it leads to significant performance improvements. Comprehensive experimental results in various conditional generation tasks validate the superiority and adaptability of DSG in terms of both sample quality and time efficiency.

The polygon mesh representation of 3D data exhibits great flexibility, fast rendering speed, and storage efficiency, which is widely preferred in various applications. However, given its unstructured graph representation, the direct generation of high-fidelity 3D meshes is challenging. Fortunately, with a pre-defined ordering strategy, 3D meshes can be represented as sequences, and the generation process can be seamlessly treated as an auto-regressive problem. In this paper, we validate Neural Coordinate Field (NeurCF), an explicit coordinate representation with implicit neural embeddings, is a simple-yet-effective representation for large-scale sequential mesh modeling. After that, we present MeshXL, a family of generative pre-trained auto-regressive models that addresses 3D mesh generation with modern large language model approaches. Extensive experiments show that MeshXL is able to generate high-quality 3D meshes, and can also serve as foundation models for various down-stream applications.

Diffusion models have garnered widespread attention in Reinforcement Learning (RL) for their powerful expressiveness and multimodality. It has been verified that utilizing diffusion policies can significantly improve the performance of RL algorithms in continuous control tasks by overcoming the limitations of unimodal policies, such as Gaussian policies. Furthermore, the multimodality of diffusion policies also shows the potential of providing the agent with enhanced exploration capabilities. However, existing works mainly focus on applying diffusion policies in offline RL, while their incorporation into online RL has been less investigated. The diffusion model's training objective, known as the variational lower bound, cannot be applied directly in online RL due to the unavailability of 'good' samples (actions). To harmonize the diffusion model with online RL, we propose a novel model-free diffusion-based online RL algorithm named Q-weighted Variational Policy Optimization (QVPO). Specifically, we introduce the Q-weighted variational loss and its approximate implementation in practice. Notably, this loss is shown to be a tight lower bound of the policy objective. To further enhance the exploration capability of the diffusion policy, we design a special entropy regularization term. Unlike Gaussian policies, the log-likelihood in diffusion policies is inaccessible; thus this entropy term is nontrivial. Moreover, to reduce the large variance of diffusion policies, we also develop an efficient behavior policy through action selection. This can further improve its sample efficiency during online interaction. Consequently, the QVPO algorithm leverages the exploration capabilities and multimodality of diffusion policies, preventing the RL agent from converging to a sub-optimal policy. To verify the effectiveness of QVPO, we conduct comprehensive experiments on MuJoCo continuous control benchmarks. The final results demonstrate that QVPO achieves state-of-the-art performance in terms of both cumulative reward and sample efficiency.

Foundation models in computer vision have demonstrated exceptional performance in zero-shot and few-shot tasks by extracting multi-purpose features from large-scale datasets through self-supervised pre-training methods. However, these models often overlook the severe corruption in cryogenic electron microscopy (cryo-EM) images by high-level noises. We introduce DRACO, a Denoising-Reconstruction Autoencoder for CryO-EM, inspired by the Noise2Noise (N2N) approach. By processing cryo-EM movies into odd and even images and treating them as independent noisy observations, we apply a denoising-reconstruction hybrid training scheme. We mask both images to create denoising and reconstruction tasks. For DRACO's pre-training, the quality of the dataset is essential, we hence build a high-quality, diverse dataset from an uncurated public database, including over 270,000 movies or micrographs. After pre-training, DRACO naturally serves as a generalizable cryo-EM image denoiser and a foundation model for various cryo-EM downstream tasks. DRACO demonstrates the best performance in denoising, micrograph curation, and particle picking tasks compared to state-of-the-art baselines.

In the past decade, deep conditional generative models have revolutionized the generation of realistic images, extending their application from entertainment to scientific domains. Single-particle cryo-electron microscopy (cryo-EM) is crucial in resolving near-atomic resolution 3D structures of proteins, such as the SARS-COV-2 spike protein. To achieve high-resolution reconstruction, a comprehensive data processing pipeline has been adopted. However, its performance is still limited as it lacks high-quality annotated datasets for training. To address this, we introduce physics-informed generative cryo-electron microscopy (CryoGEM), which for the first time integrates physics-based cryo-EM simulation with a generative unpaired noise translation to generate physically correct synthetic cryo-EM datasets with realistic noises. Initially, CryoGEM simulates the cryo-EM imaging process based on a virtual specimen. To generate realistic noises, we leverage an unpaired noise translation via contrastive learning with a novel mask-guided sampling scheme. Extensive experiments show that CryoGEM is capable of generating authentic cryo-EM images. The generated dataset can be used as training data for particle picking and pose estimation models, eventually improving the reconstruction resolution.

With the rapid advance of the Internet of Things (IoT), more and more wearable devices are being developed for real-time monitoring. Most of these existing monitors are powered by chemical batteries. Replacing and disposing batteries for an exponentially increasing number of IoT nodes prohibitively results in labor-intensive maintenance. It is also environmentally unfriendly. Gls EH, reclaiming the wasted ambient energy, is a promising technology for battery-free IoT. This article presents a novel motion-powered real-time human activity recognition (HAR) system called motion-powered HAR system (MP-HAR), where the harvester works as both an energy source and sensor. MP-HAR emphasizes low-power as well as low-cost characteristics, encompassing four necessary units: 1) energy transduction unit (ETU); 2) energy management unit (EMU); 3) energy user unit (EUU); and 4) edge computing unit (ECU). In particular, the unique intermittent operation based on the reconfigurable on/off threshold voltages given by the well-rounded energy-aware circuit has been discussed in detail. The balance between energy supply and information demand in MP-HAR has been achieved by using a handy design. Utilizing the unique correspondence between human arm swing frequency and harvested energy, the information flows with energy inside the system. By knowing the interval between transmitted packets, MP-HAR has realized HAR in real time. Moreover, an all-in-one prototype has been fabricated to validate the performance of the proposed system. Lab and field tests have demonstrated that MP-HAR can reliably recognize different human activities, such as standing, walking, jogging, and running. As a cyber-electro-mechanical co-design, MP-HAR has brought a promising solution for pervasive HAR and ubiquitous IoT.

Neural Volume Rendering (NVR) has advanced explosively since the advent of Neural Radiance Field (NeRF), a technique for novel view synthesis of complex scenes based on a finite set of input views. Existing ray casting-based NVR approaches process rays concurrently to leverage parallelism but fails to consider its impact on cache locality, which ultimately undermines the efficiency of corresponding dedicated hardware accelerator designs. We further observed that there exhibits spatial correspondence between features and voxels in NVR that can be exploited by processing in the order of voxel, not ray. This paper introduces a novel approach to meticulously reorder the execution of rays, ensuring that rays with similar memory access patterns are processed in parallel, thereby enhancing cache locality. On the basis of that, we also propose an efficient backend architecture and a corresponding memory subsystem, facilitating accurate data prefetching to hide off-chip memory latency. To validate the proposed architecture, we implement our design in VerilogHDL and evaluate the performance by post-synthesis simulation with real scene data. The evaluation results demonstrate that our design markedly enhances the efficiency of NVR processing, achieving a considerable speedup ($1.62\times $) compared to the state-of-the-art NVR accelerator, while necessitating significantly less silicon area ($5.12\times $) and power ($32.79\times $).

Large garages are ubiquitous yet intricate scenes that present unique challenges due to their monotonous colors, repetitive patterns, reflective surfaces, and transparent vehicle glass. Conventional Structure from Motion (SfM) methods for camera pose estimation and 3D reconstruction often fail in these environments due to poor correspondence construction. To address these challenges, we introduce LetsGo, a LiDAR-assisted Gaussian splatting framework for large-scale garage modeling and rendering. We develop a handheld scanner, Polar, equipped with IMU, LiDAR, and a fisheye camera, to facilitate accurate data acquisition. Using this Polar device, we present the GarageWorld dataset, consisting of eight expansive garage scenes with diverse geometric structures, which will be made publicly available for further research. Our approach demonstrates that LiDAR point clouds collected by the Polar device significantly enhance a suite of 3D Gaussian splatting algorithms for garage scene modeling and rendering. We introduce a novel depth regularizer that effectively eliminates floating artifacts in rendered images. Additionally, we propose a multi-resolution 3D Gaussian representation designed for Level-of-Detail (LOD) rendering. This includes adapted scaling factors for individual levels and a random-resolution-level training scheme to optimize the Gaussians across different resolutions. This representation enables efficient rendering of large-scale garage scenes on lightweight devices via a web-based renderer. Experimental results on our GarageWorld dataset, as well as on ScanNet++ and KITTI-360, demonstrate the superiority of our method in terms of rendering quality and resource efficiency.

Apparel's significant role in human appearance underscores the importance of garment digitalization for digital human creation. Recent advances in 3D content creation are pivotal for digital human creation. Nonetheless, garment generation from text guidance is still nascent. We introduce a text-driven 3D garment generation framework, DressCode, which aims to democratize design for novices and offer immense potential in fashion design, virtual try-on, and digital human creation. We first introduce SewingGPT, a GPT-based architecture integrating cross-attention with text-conditioned embedding to generate sewing patterns with text guidance. We then tailor a pre-trained Stable Diffusion to generate tile-based Physically-based Rendering (PBR) textures for the garments. By leveraging a large language model, our framework generates CG-friendly garments through natural language interaction. It also facilitates pattern completion and texture editing, streamlining the design process through user-friendly interaction. This framework fosters innovation by allowing creators to freely experiment with designs and incorporate unique elements into their work. With comprehensive evaluations and comparisons with other state-of-the-art methods, our method showcases superior quality and alignment with input prompts. User studies further validate our high-quality rendering results, highlighting its practical utility and potential in production settings. Our project page is https://IHe-KaiI.github.io/DressCode/.

Volumetric video represents a transformative advancement in visual media, enabling users to freely navigate immersive virtual experiences and narrowing the gap between digital and real worlds. However, the need for extensive manual intervention to stabilize mesh sequences and the generation of excessively large assets in existing workflows impedes broader adoption. In this paper, we present a novel Gaussian-based approach, dubbed DualGS , for real-time and high-fidelity playback of complex human performance with excellent compression ratios. Our key idea in DualGS is to separately represent motion and appearance using the corresponding skin and joint Gaussians. Such an explicit disentanglement can significantly reduce motion redundancy and enhance temporal coherence. We begin by initializing the DualGS and anchoring skin Gaussians to joint Gaussians at the first frame. Subsequently, we employ a coarse-to-fine training strategy for frame-by-frame human performance modeling. It includes a coarse alignment phase for overall motion prediction as well as a fine-grained optimization for robust tracking and high-fidelity rendering. To integrate volumetric video seamlessly into VR environments, we efficiently compress motion using entropy encoding and appearance using codec compression coupled with a persistent codebook. Our approach achieves a compression ratio of up to 120 times, only requiring approximately 350KB of storage per frame. We demonstrate the efficacy of our representation through photo-realistic, free-view experiences on VR headsets, enabling users to immersively watch musicians in performance and feel the rhythm of the notes at the performers' fingertips. Project page: https://nowheretrix.github.io/DualGS/.

Experiencing high-fidelity volumetric video as seamlessly as 2D videos is a long-held dream. However, current dynamic 3DGS methods, despite their high rendering quality, face challenges in streaming on mobile devices due to computational and bandwidth constraints. In this paper, we introduce V 3 (Viewing Volumetric Videos), a novel approach that enables high-quality mobile rendering through the streaming of dynamic Gaussians. Our key innovation is to view dynamic 3DGS as 2D videos, facilitating the use of hardware video codecs. Additionally, we propose a two-stage training strategy to reduce storage requirements with rapid training speed. The first stage employs hash encoding and shallow MLP to learn motion, then reduces the number of Gaussians through pruning to meet the streaming requirements, while the second stage fine tunes other Gaussian attributes using residual entropy loss and temporal loss to improve temporal continuity. This strategy, which disentangles motion and appearance, maintains high rendering quality with compact storage requirements. Meanwhile, we designed a multi-platform player to decode and render 2D Gaussian videos. Extensive experiments demonstrate the effectiveness of V 3 , outperforming other methods by enabling high-quality rendering and streaming on common devices, which is unseen before. As the first to stream dynamic Gaussians on mobile devices, our companion player offers users an unprecedented volumetric video experience, including smooth scrolling and instant sharing. Our project page with source code is available at https://authoritywang.github.io/v3/.

In the field of trajectory generation for objects, ensuring continuous collision-free motion remains a huge challenge, especially for non-convex geometries and complex environments. Previous methods either oversimplify object shapes, which results in a sacrifice of feasible space or rely on discrete sampling, which suffers from the "tunnel effect". To address these limitations, we propose a novel hierarchical trajectory generation pipeline, which utilizes the Swept Volume Signed Distance Field (SVSDF) to guide trajectory optimization for Continuous Collision Avoidance (CCA). Our interdisciplinary approach, blending techniques from graphics and robotics, exhibits outstanding effectiveness in solving this problem. We formulate the computation of the SVSDF as a Generalized Semi-Infinite Programming model, and we solve for the numerical solutions at query points implicitly, thereby eliminating the need for explicit reconstruction of the surface. Our algorithm has been validated in a variety of complex scenarios and applies to robots of various dynamics, including both rigid and deformable shapes. It demonstrates exceptional universality and superior CCA performance compared to typical algorithms. The code will be released at https://github.com/ZJU-FAST-Lab/Implicit-SVSDF-Planner for the benefit of the community.

In the realm of digital creativity, our potential to craft intricate 3D worlds from imagination is often hampered by the limitations of existing digital tools, which demand extensive expertise and efforts. To narrow this disparity, we introduce CLAY, a 3D geometry and material generator designed to effortlessly transform human imagination into intricate 3D digital structures. CLAY supports classic text or image inputs as well as 3D-aware controls from diverse primitives (multi-view images, voxels, bounding boxes, point clouds, implicit representations, etc). At its core is a large-scale generative model composed of a multi-resolution Variational Autoencoder (VAE) and a minimalistic latent Diffusion Transformer (DiT), to extract rich 3D priors directly from a diverse range of 3D geometries. Specifically, it adopts neural fields to represent continuous and complete surfaces and uses a geometry generative module with pure transformer blocks in latent space. We present a progressive training scheme to train CLAY on an ultra large 3D model dataset obtained through a carefully designed processing pipeline, resulting in a 3D native geometry generator with 1.5 billion parameters. For appearance generation, CLAY sets out to produce physically-based rendering (PBR) textures by employing a multi-view material diffusion model that can generate 2K resolution textures with diffuse, roughness, and metallic modalities. We demonstrate using CLAY for a range of controllable 3D asset creations, from sketchy conceptual designs to production ready assets with intricate details. Even first time users can easily use CLAY to bring their vivid 3D imaginations to life, unleashing unlimited creativity.

Depth estimation is usually ill-posed and ambiguous for monocular camera-based 3D multi-person pose estimation. Since LiDAR can capture accurate depth information in long-range scenes, it can benefit both the global localization of individuals and the 3D pose estimation by providing rich geometry features. Motivated by this, we propose a monocular camera and single LiDAR-based method for 3D multi-person pose estimation in large-scale scenes, which is easy to deploy and insensitive to light. Specifically, we design an effective fusion strategy to take advantage of multi-modal input data, including images and point cloud, and make full use of temporal information to guide the network to learn natural and coherent human motions. Without relying on any 3D pose annotations, our method exploits the inherent geometry constraints of point cloud for self-supervision and utilizes 2D keypoints on images for weak supervision. Extensive experiments on public datasets and our newly collected dataset demonstrate the superiority and generalization capability of our proposed method. Project homepage is at \url{https://github.com/4DVLab/FusionPose.git}.

Monocular 3D motion capture (mocap) is beneficial to many applications. The use of a single camera, however, often fails to handle occlusions of different body parts and hence it is limited to capture relatively simple movements. We present a light-weight, hybrid mocap technique called HybridCap that augments the camera with only 4 Inertial Measurement Units (IMUs) in a novel learning-and-optimization framework. We first employ a weakly-supervised and hierarchical motion inference module based on cooperative pure residual recurrent blocks that serve as limb, body and root trackers as well as an inverse kinematics solver. Our network effectively narrows the search space of plausible motions via coarse-to-fine pose estimation and manages to tackle challenging movements with high efficiency. We further develop a hybrid optimization scheme that combines inertial feedback and visual cues to improve tracking accuracy. Extensive experiments on various datasets demonstrate HybridCap can robustly handle challenging movements ranging from fitness actions to Latin dance. It also achieves real-time performance up to 60 fps with state-of-the-art accuracy.

This paper presents an inverse kinematic optimization layer (IKOL) for 3D human pose and shape estimation that leverages the strength of both optimization- and regression-based methods within an end-to-end framework. IKOL involves a nonconvex optimization that establishes an implicit mapping from an image’s 3D keypoints and body shapes to the relative body-part rotations. The 3D keypoints and the body shapes are the inputs and the relative body-part rotations are the solutions. However, this procedure is implicit and hard to make differentiable. So, to overcome this issue, we designed a Gauss-Newton differentiation (GN-Diff) procedure to differentiate IKOL. GN-Diff iteratively linearizes the nonconvex objective function to obtain Gauss-Newton directions with closed form solutions. Then, an automatic differentiation procedure is directly applied to generate a Jacobian matrix for end-to-end training. Notably, the GN-Diff procedure works fast because it does not rely on a time-consuming implicit differentiation procedure. The twist rotation and shape parameters are learned from the neural networks and, as a result, IKOL has a much lower computational overhead than most existing optimization-based methods. Additionally, compared to existing regression-based methods, IKOL provides a more accurate mesh-image correspondence. This is because it iteratively reduces the distance between the keypoints and also enhances the reliability of the pose structures. Extensive experiments demonstrate the superiority of our proposed framework over a wide range of 3D human pose and shape estimation methods. Code is available at https://github.com/Juzezhang/IKOL

The success of the Neural Radiance Fields (NeRFs) for modeling and free-view rendering static objects has in-spired numerous attempts on dynamic scenes. Current techniques that utilize neural rendering for facilitating free-view videos (FVVs) are restricted to either offline rendering or are capable of processing only brief sequences with minimal motion. In this paper, we present a novel technique, Residual Radiance Field or ReRF, as a highly com-pact neural representation to achieve real-time FVV ren-dering on long-duration dynamic scenes. ReRF explicitly models the residual information between adjacent times-tamps in the spatial-temporal feature space, with a global coordinate-based tiny MLP as the feature decoder. Specif-ically, ReRF employs a compact motion grid along with a residual feature grid to exploit inter-frame feature similar-ities. We show such a strategy can handle large motions without sacrificing quality. We further present a sequential training scheme to maintain the smoothness and the spar-sity of the motion/residual grids. Based on ReRF, we design a special FVV codec that achieves three orders of magni-tudes compression rate and provides a companion ReRF player to support online streaming of long-duration FVVs of dynamic scenes. Extensive experiments demonstrate the effectiveness of ReRF for compactly representing dynamic radiance fields, enabling an unprecedented free-viewpoint viewing experience in speed and quality.

Humans constantly interact with objects in daily life tasks. Capturing such processes and subsequently conducting visual inferences from a fixed viewpoint suffers from occlusions, shape and texture ambiguities, motions, etc. To mitigate the problem, it is essential to build a training dataset that captures free-viewpoint interactions. We construct a dense multi-view dome to acquire a complex human object interaction dataset, named HODome, that consists of ~71 M frames on 10 subjects interacting with 23 objects. To process the HODome dataset, we develop NeuralDome, a layer-wise neural processing pipeline tailored for multi-view video inputs to conduct accurate tracking, geometry reconstruction and free-view rendering, for both human subjects and objects. Extensive experiments on the HODome dataset demonstrate the effectiveness of NeuralDome on a variety of inference, modeling, and rendering tasks. Both the dataset and the NeuralDome tools will be disseminated to the community for further development, which can be found at https://juzezhang.github.io/NeuralDome

Human modeling and relighting are two fundamental problems in computer vision and graphics, where high-quality datasets can largely facilitate related research. However, most existing human datasets only provide multi-view human images captured under the same illumination. Although valuable for modeling tasks, they are not read-ily used in relighting problems. To promote research in both fields, in this paper, we present UltraStage, a new 3D human dataset that contains more than 2, 000 high-quality human assets captured under both multi-view and multi-illumination settings. Specifically, for each example, we provide 32 surrounding views illuminated with one white light and two gradient illuminations. In addition to regular multi-view images, gradient illuminations help recover de-tailed surface normal and spatially-varying material maps, enabling various relighting applications. Inspired by recent advances in neural representation, we further interpret each example into a neural human asset which allows novel view synthesis under arbitrary lighting conditions. We show our neural human assets can achieve extremely high capture performance and are capable of representing fine details such as facial wrinkles and cloth folds. We also validate UltraStage in single image relighting tasks, training neural networks with virtual relighted data from neural assets and demonstrating realistic rendering improvements over prior arts. UltraStage will be publicly available to the community to stimulate significant future developments in various human modeling and rendering tasks. The dataset is available at https://miaoing.github.io/RNHA.

Reconstructing an object behind a scattering medium needs to tackle different light scatterings inside the medium and in the free space. Major approaches, e.g., diffuse optical tomography and non-line-of-sight imaging, address either of the light scatterings. Confocal diffuse tomography (CDT) considers the media acting as a diffuse kernel on the free-space scattering and recovers the object by deconvolving the inside-medium scattering. Inspired by CDT, we present a Neural De-Scatterer to solve this challenging problem. We exploit neural implicit fields to represent the the free-space scattering and use a multilayer perceptron (MLP) to learn the density and albedo of the hidden object. Furthermore, we tailor a bi-directional training strategy to optimize the MLP with forward and backward losses and employ hash encoding for memory and computation efficiency. The Neural De-Scatterer enables us to reconstruct objects at arbitrary resolution. Comprehensive experiments with synthetic and real measurements demonstrate that our Neural De-Scatterer outperforms state-of-the-art methods. Our data and code are publicly available.

We present a novel Neural Refractive Field (NeReF) to recover wavefront of transparent fluids by simultaneously estimating the surface position and normal of the fluid front. Unlike prior arts that treat the reconstruction target as a single layer of the surface, NeReF is specifically formulated to recover a volumetric normal field with its corresponding density field. A query ray will be refracted by NeReF according to its accumulated refractive point and normal, and we employ the correspondences and uniqueness of refracted ray for NeReF optimization. We show NeReF, as a global optimization scheme, can more robustly tackle refraction distortions detrimental to traditional methods for correspondence matching. Furthermore, the continuous NeReF representation of wavefront enables view synthesis as well as normal integration. We validate our approach on both synthetic and real data and show it is particularly suitable for sparse multi-view acquisition. We hence build a small light field array and experiment on various surface shapes to demonstrate high fidelity NeReF reconstruction.

We present a novel type of neural fields that uses general radial bases for signal representation. State-of-the-art neural fields typically rely on grid-based representations for storing local neural features and N-dimensional linear kernels for interpolating features at continuous query points. The spatial positions of their neural features are fixed on grid nodes and cannot well adapt to target signals. Our method instead builds upon general radial bases with flexible kernel position and shape, which have higher spatial adaptivity and can more closely fit target signals. To further improve the channel-wise capacity of radial basis functions, we propose to compose them with multi-frequency sinusoid functions. This technique extends a radial basis to multiple Fourier radial bases of different frequency bands without requiring extra parameters, facilitating the representation of details. Moreover, by marrying adaptive radial bases with grid-based ones, our hybrid combination inherits both adaptivity and interpolation smoothness. We carefully designed weighting schemes to let radial bases adapt to different types of signals effectively. Our experiments on 2D image and 3D signed distance field representation demonstrate the higher accuracy and compactness of our method than prior arts. When applied to neural radiance field reconstruction, our method achieves state-of-the-art rendering quality, with small model size and comparable training speed.

Task driven object detection aims to detect object instances suitable for affording a task in an image. Its challenge lies in object categories available for the task being too diverse to be limited to a closed set of object vocabulary for traditional object detection. Simply mapping categories and visual features of common objects to the task cannot address the challenge. In this paper, we propose to explore fundamental affordances rather than object categories, i.e., common attributes that enable different objects to accomplish the same task. Moreover, we propose a novel multi-level chain-of-thought prompting (MLCoT) to extract the affordance knowledge from large language models, which contains multi-level reasoning steps from task to object examples to essential visual attributes with rationales. Furthermore, to fully exploit knowledge to benefit object recognition and localization, we propose a knowledge-conditional detection framework, namely CoTDet. It conditions the detector from the knowledge to generate object queries and regress boxes. Experimental results demonstrate that our CoTDet outperforms state-of-the-art methods consistently and significantly (+15.6 box AP and +14.8 mask AP) and can generate rationales for why objects are detected to afford the task.

Human-centric scene understanding is significant for real-world applications, but it is extremely challenging due to the existence of diverse human poses and actions, complex human-environment interactions, severe occlusions in crowds, etc. In this paper, we present a large-scale multi-modal dataset for human-centric scene under-standing, dubbed HuCenLife, which is collected in diverse daily-life scenarios with rich and fine-grained annotations. Our HuCenLife can benefit many 3D perception tasks, such as segmentation, detection, action recognition, etc., and we also provide benchmarks for these tasks to facilitate related research. In addition, we design novel modules for LiDAR-based segmentation and action recognition, which are more applicable for large-scale human-centric scenarios and achieve state-of-the-art performance. The dataset and code can be found at https://github.com/4DVLab/HuCenLife.git.

Recovering the physical attributes of an object’s appearance from its images captured under an unknown illumination is challenging yet essential for photo-realistic rendering. Recent approaches adopt the emerging implicit scene representations and have shown impressive results. However, they unanimously adopt a surface-based representation, and hence can not well handle scenes with very complex geometry, translucent object and etc. In this paper, we propose to conduct inverse volume rendering, in contrast to surface-based, by representing a scene using microflake volume, which assumes the space is filled with infinite small flakes and light reflects or scatters at each spatial location according to microflake distributions. We further adopt the coordinate networks to implicitly encode the microflake volume, and develop a differentiable microflake volume renderer to train the network in an end-to-end way in principle. Our NeMF enables effective recovery of appearance attributes for highly complex geometry and scattering object, enables high-quality relighting, material editing, and especially simulates volume rendering effects, such as scattering, which is infeasible for surface-based approaches. Our data and code are available at: https://github.com/YoujiaZhang/NeMF.

A gigapixel image consists of billions of pixels with color information to record fine details of the scene, leading to tremendous data overload for storage and display. Recent advances of gigapixel imaging still suffer from large storage size, I/O overhead or spatial aliasing for achieving real-time rendering especially during zoom-in or zoom-out. To fill this gap, in this paper, we propose NeuralGiga, a novel neural representation of gigapixel images with an effective neural rendering scheme. NeuralGiga implicitly encodes the entire image into a light-weight network which maps pixel coordinates into RGB values with efficient storage overload. In our novel neural rendering network, to enable high-quality giga-image regression with anti-aliasing and continuous viewing effect, we introduce a Spectrum Multi-Layer Perceptron (MLP) design and a Gaussian-based Integrated Random Fourier Feature Mapping (GIRFFM) scheme. Extensive experiments on various scenarios illustrate the effectiveness of our approach to achieve high-quality neural giga-image representation for both storage and display.

We present a novel multi-view dynamic 3D human reconstruction technique based on model-based shape deformation. Our approach specifically targets at handling challenging cases such as textureless appearance, heavy occlusions, and depth order ambiguity that are problematic to stereo-based techniques. We propose to pose match and shape deform a human template model to avoid meshing the point cloud. To robustly match the template pose with image observations, we present a novel Graph Convolutional Networks (GCN) to gradually filter out erroneous views and impose appropriate weights on the optimal subset for recovering the 3D skeleton and warping the template shape. Next, We use the warped human template to guide the cross-view consistent semantic segmentation. We set out to deform the warped 3D model so that the silhouette of the deformed model best matches the target in respective views while maintaining semantic consistency. Comprehensive experiments on publicly available and our newly generated complex motion datasets show our approach significantly outperforms the state-of-the-art on sparse cameras, textureless regions (e.g., under black clothing), complex motions, etc.

Modeling and capturing the 3D spatial arrangement of the human and the object is the key to perceiving 3D human-object interaction from monocular images. In this work, we propose to use the Human-Object Offset between anchors which are densely sampled from the surface of human mesh and object mesh to represent human-object spatial relation. Compared with previous works which use contact map or implicit distance filed to encode 3D human-object spatial relations, our method is a simple and efficient way to encode the highly detailed spatial correlation between the human and object. Based on this representation, we propose Stacked Normalizing Flow (StackFLOW) to infer the posterior distribution of human-object spatial relations from the image. During the optimization stage, we finetune the human body pose and object 6D pose by maximizing the likelihood of samples based on this posterior distribution and minimizing the 2D-3D corresponding reprojection loss. Extensive experimental results show that our method achieves impressive results on two challenging benchmarks, BEHAVE and InterCap datasets. Our code has been publicly available at https://github.com/MoChen-bop/StackFLOW.

Text-to-video is a rapidly growing research area that aims to generate a semantic, identical, and temporal coherence sequence of frames that accurately align with the input text prompt. This study focuses on zero-shot text-to-video generation considering the data- and cost-efficient. To generate a semantic-coherent video, exhibiting a rich portrayal of temporal semantics such as the whole process of flower blooming rather than a set of ``moving images'', we propose a novel Free-Bloom pipeline that harnesses large language models (LLMs) as the director to generate a semantic-coherence prompt sequence, while pre-trained latent diffusion models (LDMs) as the animator to generate the high fidelity frames. Furthermore, to ensure temporal and identical coherence while maintaining semantic coherence, we propose a series of annotative modifications to adapting LDMs in the reverse process, including joint noise sampling, step-aware attention shift, and dual-path interpolation. Without any video data and training requirements, Free-Bloom generates vivid and high-quality videos, awe-inspiring in generating complex scenes with semantic meaningful frame sequences. In addition, Free-Bloom is naturally compatible with LDMs-based extensions.

Though the advancement of pre-trained large language models unfolds, the exploration of building a unified model for language and other multimodal data, such as motion, remains challenging and untouched so far. Fortunately, human motion displays a semantic coupling akin to human language, often perceived as a form of body language. By fusing language data with large-scale motion models, motion-language pre-training that can enhance the performance of motion-related tasks becomes feasible. Driven by this insight, we propose MotionGPT, a unified, versatile, and user-friendly motion-language model to handle multiple motion-relevant tasks. Specifically, we employ the discrete vector quantization for human motion and transfer 3D motion into motion tokens, similar to the generation process of word tokens. Building upon this "motion vocabulary", we perform language modeling on both motion and text in a unified manner, treating human motion as a specific language. Moreover, inspired by prompt learning, we pre-train MotionGPT with a mixture of motion-language data and fine-tune it on prompt-based question-and-answer tasks. Extensive experiments demonstrate that MotionGPT achieves state-of-the-art performances on multiple motion tasks including text-driven motion generation, motion captioning, motion prediction, and motion in-between.

Emerging neural reconstruction techniques based on tomography (e.g., NeRF, NeAT, and NeRP) have started showing unique capabilities in medical imaging. In this work, we present a novel Polychromatic neural representation (Polyner) to tackle the challenging problem of CT imaging when metallic implants exist within the human body. CT metal artifacts arise from the drastic variation of metal's attenuation coefficients at various energy levels of the X-ray spectrum, leading to a nonlinear metal effect in CT measurements. Recovering CT images from metal-affected measurements hence poses a complicated nonlinear inverse problem where empirical models adopted in previous metal artifact reduction (MAR) approaches lead to signal loss and strongly aliased reconstructions. Polyner instead models the MAR problem from a nonlinear inverse problem perspective. Specifically, we first derive a polychromatic forward model to accurately simulate the nonlinear CT acquisition process. Then, we incorporate our forward model into the implicit neural representation to accomplish reconstruction. Lastly, we adopt a regularizer to preserve the physical properties of the CT images across different energy levels while effectively constraining the solution space. Our Polyner is an unsupervised method and does not require any external training data. Experimenting with multiple datasets shows that our Polyner achieves comparable or better performance than supervised methods on in-domain datasets while demonstrating significant performance improvements on out-of-domain datasets. To the best of our knowledge, our Polyner is the first unsupervised MAR method that outperforms its supervised counterparts. The code for this work is available at: https://github.com/iwuqing/Polyner.

Abstract Neural Radiance Fields (NeRF) have significantly advanced the generation of highly realistic and expressive 3D scenes. However, the task of editing NeRF, particularly in terms of geometry modification, poses a significant challenge. This issue has obstructed NeRF's wider adoption across various applications. To tackle the problem of efficiently editing neural implicit fields, we introduceNeural Impostor, a hybrid representation incorporating an explicit tetrahedral mesh alongside a multigrid implicit field designated for each tetrahedron within the explicit mesh. Our framework bridges the explicit shape manipulation and the geometric editing of implicit fields by utilizing multigrid barycentric coordinate encoding, thus offering a pragmatic solution to deform, composite, and generate neural implicit fields while maintaining a complex volumetric appearance. Furthermore, we propose a comprehensive pipeline for editing neural implicit fields based on a set of explicit geometric editing operations. We show the robustness and adaptability of our system through diverse examples and experiments, including the editing of both synthetic objects and real captured data. Finally, we demonstrate the authoring process of a hybrid synthetic‐captured object utilizing a variety of editing operations, underlining the transformative potential ofNeural Impostorin the field of 3D content creation and manipulation.

Abstract The development of neural relighting techniques has by far outpaced the rate of their corresponding training data (e.g., OLAT) generation. For example, high-quality relighting from a single portrait image still requires supervision from comprehensive datasets covering broad diversities in gender, race, complexion, and facial geometry. We present a hybrid parametric neural relighting (PN-Relighting) framework for single portrait relighting, using a much smaller OLAT dataset or SMOLAT. At the core of PN-Relighting, we employ parametric 3D faces coupled with appearance inference and implicit material modelling to enrich SMOLAT for handling in-the-wild images. Specifically, we tailor an appearance inference module to generate detailed geometry and albedo on top of the parametric face and develop a neural rendering module to first construct an implicit material representation from SMOLAT and then conduct self-supervised training on in-the-wild image datasets. Comprehensive experiments show that PN-Relighting produces comparable high-quality relighting to TotalRelighting (Pandey et al., 2021), but with a smaller dataset. It further improves shape estimation and naturally supports free-viewpoint rendering and partial skin material editing. PN-Relighting also serves as a data augmenter to produce rich OLAT datasets beyond the original capture.

Light field (LF) cameras record both intensity and directions of light rays, and encode 3D scenes into 4D LF images. Recently, many convolutional neural networks (CNNs) have been proposed for various LF image processing tasks. However, it is challenging for CNNs to effectively process LF images since the spatial and angular information are highly inter-twined with varying disparities. In this paper, we propose a generic mechanism to disentangle these coupled information for LF image processing. Specifically, we first design a class of domain-specific convolutions to disentangle LFs from different dimensions, and then leverage these disentangled features by designing task-specific modules. Our disentangling mechanism can well incorporate the LF structure prior and effectively handle 4D LF data. Based on the proposed mechanism, we develop three networks (i.e., DistgSSR, DistgASR and DistgDisp) for spatial super-resolution, angular super-resolution and disparity estimation. Experimental results show that our networks achieve state-of-the-art performance on all these three tasks, which demonstrates the effectiveness, efficiency, and generality of our disentangling mechanism. Project page: https://yingqianwang.github.io/DistgLF/.

Particle Imaging Velocimetry (PIV) is a classical method that estimates fluid flow by analyzing the motion of injected particles. To reconstruct and track the swirling particles is a difficult computer vision problem, as the particles are dense in the fluid volume and have similar appearances. Further, tracking a large number of particles is particularly challenging due to heavy occlusion. Here we present a low-cost PIV solution that uses compact lenslet-based light field cameras as imaging device. We develop novel optimization algorithms for dense particle 3D reconstruction and tracking. As a single light field camera has limited capacity in resolving depth (z-dimension measurement), the resolution of 3D reconstruction on the x-y plane is much higher than along the z-axis. To compensate for the imbalanced resolution in 3D, we use two light field cameras positioned at an orthogonal angle to capture particle images. In this way, we can achieve high-resolution 3D particle reconstruction in the full fluid volume. For each time frame, we first estimate particle depths under a single viewpoint by exploiting the focal stack symmetry of light field. We then fuse the recovered 3D particles in two views by solving a linear assignment problem (LAP). Specifically, we propose an anisotropic point-to-ray distance as matching cost to handle the resolution mismatch. Finally, given a sequence of 3D particle reconstructions over time, we recover the full-volume 3D fluid flow with a physically-constrained optical flow, which enforces local motion rigidity and fluid incompressibility. We perform comprehensive experiments on synthetic and real data for ablation and evaluation. We show that our method recovers full-volume 3D fluid flows of various types. Two-view reconstruction results achieves higher accuracy than those with one view only.

Neural Radiance Field (NeRF), a disruptive method for 3D representation and rendering, is extremely popular in the field of computer graphics and computer vision in the past three years. The most distinctive feature of NeRF models is their scene representation property, making it possible to quantize the models according to the complexity of the representing scenes. This paper proposes a novel approach to improve the efficiency of NeRF rendering by adopting precision-scalable computation. We first analyze and validate the idea of scene-dependent quantization for NeRF models. Based on that, we further propose look-up table (LUT) processing element (PE)-based precision-scalable computation unit designs. To evaluate the performance of different precision-scalable computing units, we implement these designs and compare the corresponding area, power, speed and energy efficiency. We also compare the proposed designs with existing approaches as well as the fixed precision approach for NeRF rendering tasks. The comparison results show that energy efficiency can be significantly improved by using precision-scalable computation for NeRF.

This work presents an energy-efficient accelerator for medical image reconstruction from implicit neural representation (INR). The accelerator implements an INR-based algorithm to deliver high-quality medical image reconstruction with arbitrary resolution from a compact implicit format. In particular, we propose a dedicated hardware architecture based on an optimized computation flow for the INR-based reconstruction algorithm, which co-designs data reuse and computation load. The proposed architecture takes in the coordinate of the intersection of three scans and outputs all the voxel intensities, minimizing the data movement between on-chip and off-chip. To validate the proposed accelerator, we build a proof-of-concept prototype demonstration system using field programmable gate array (FPGA). We also map our design to 40nm CMOS technology to measure the performance of the proposed accelerator. The implementation results show that, running at 400MHz, the proposed accelerator is capable of processing medical images with$256\times 256$resolution in real-time at 26.3 frames per second (FPS), with a power consumption of only 795 mW. Comparison results show that the performance, as well as the energy efficiency of the proposed accelerator, outperforms the central processing unit (CPU)-based and graphic processing unit (GPU)-based implementations.

High Resolution (HR) medical images provide rich anatomical structure details to facilitate early and accurate diagnosis. In magnetic resonance imaging (MRI), restricted by hardware capacity, scan time, and patient cooperation ability, isotropic 3-dimensional (3D) HR image acquisition typically requests long scan time and, results in small spatial coverage and low signal-to-noise ratio (SNR). Recent studies showed that, with deep convolutional neural networks, isotropic HR MR images could be recovered from low-resolution (LR) input via single image super-resolution (SISR) algorithms. However, most existing SISR methods tend to approach scale-specific projection between LR and HR images, thus these methods can only deal with fixed up-sampling rates. In this paper, we propose ArSSR, an Arbitrary Scale Super-Resolution approach for recovering 3D HR MR images. In the ArSSR model, the LR image and the HR image are represented using the same implicit neural voxel function with different sampling rates. Due to the continuity of the learned implicit function, a single ArSSR model is able to achieve arbitrary and infinite up-sampling rate reconstructions of HR images from any input LR image. Then the SR task is converted to approach the implicit voxel function via deep neural networks from a set of paired HR and LR training examples. The ArSSR model consists of an encoder network and a decoder network. Specifically, the convolutional encoder network is to extract feature maps from the LR input images and the fully-connected decoder network is to approximate the implicit voxel function. Experimental results on three datasets show that the ArSSR model can achieve state-of-the-art SR performance for 3D HR MR image reconstruction while using a single trained model to achieve arbitrary up-sampling scales.

Emerging Metaverse applications demand accessible, accurate and easy-to-use tools for 3D digital human creations in order to depict different cultures and societies as if in the physical world. Recent large-scale vision-language advances pave the way for novices to conveniently customize 3D content. However, the generated CG-friendly assets still cannot represent the desired facial traits for human characteristics. In this paper, we present Dream-Face, a progressive scheme to generate personalized 3D faces under text guidance. It enables layman users to naturally customize 3D facial assets that are compatible with CG pipelines, with desired shapes, textures and fine-grained animation capabilities. From a text input to describe the facial traits, we first introduce a coarse-to-fine scheme to generate the neutral facial geometry with a unified topology. We employ a selection strategy in the CLIP embedding space to generate coarse geometry, and subsequently optimize both the detailed displacements and normals using Score Distillation Sampling (SDS) from the generic Latent Diffusion Model (LDM). Then, for neutral appearance generation, we introduce a dual-path mechanism, which combines the generic LDM with a novel texture LDM to ensure both the diversity and textural specification in the UV space. We also employ a two-stage optimization to perform SDS in both the latent and image spaces to significantly provide compact priors for fine-grained synthesis. It also enables learning the mapping from the compact latent space into physically-based textures (diffuse albedo, specular intensity, normal maps, etc.). Our generated neutral assets naturally support blendshapes-based facial animations, thanks to the unified geometric topology. We further improve the animation ability with personalized deformation characteristics. To this end, we learn the universal expression prior in a latent space with neutral asset conditioning using the cross-identity hypernetwork, we subsequently train a neural facial tracker from video input space into the pre-trained expression space for personalized fine-grained animation. Extensive qualitative and quantitative experiments validate the effectiveness and generalizability of DreamFace. Notably, DreamFace can generate realistic 3D facial assets with physically-based rendering quality and rich animation ability from video footage, even for fashion icons or exotic characters in cartoons and fiction movies.

Significant advancements have been made in developing parametric models for digital humans, with various approaches concentrating on parts such as the human body, hand, or face. Nevertheless, connectors such as the neck have been overlooked in these models, with rich anatomical priors often unutilized. In this paper, we introduce HACK (Head-And-neCK), a novel parametric model for constructing the head and cervical region of digital humans. Our model seeks to disentangle the full spectrum of neck and larynx motions, facial expressions, and appearance variations, providing personalized and anatomically consistent controls, particularly for the neck regions. To build our HACK model, we acquire a comprehensive multi-modal dataset of the head and neck under various facial expressions. We employ a 3D ultrasound imaging scheme to extract the inner biomechanical structures, namely the precise 3D rotation information of the seven vertebrae of the cervical spine. We then adopt a multi-view photometric approach to capture the geometry and physically-based textures of diverse subjects, who exhibit a diverse range of static expressions as well as sequential head-and-neck movements. Using the multi-modal dataset, we train the parametric HACK model by separating the 3D head and neck depiction into various shape, pose, expression, and larynx blendshapes from the neutral expression and the rest skeletal pose. We adopt an anatomically-consistent skeletal design for the cervical region, and the expression is linked to facial action units for artist-friendly controls. We also propose to optimize the mapping from the identical shape space to the PCA spaces of personalized blendshapes to augment the pose and expression blendshapes, providing personalized properties within the framework of the generic model. Furthermore, we use larynx blendshapes to accurately control the larynx deformation and force the larynx slicing motions along the vertical direction in the UV-space for precise modeling of the larynx beneath the neck skin. HACK addresses the head and neck as a unified entity, offering more accurate and expressive controls, with a new level of realism, particularly for the neck regions. This approach has significant benefits for numerous applications, including geometric fitting and animation, and enables inter-correlation analysis between head and neck for fine-grained motion synthesis and transfer.

We propose a multi-sensor fusion method for capturing challenging 3D human motions with accurate consecutive local poses and global trajectories in large-scale scenarios, only using single LiDAR and 4 IMUs, which are set up conveniently and worn lightly. Specifically, to fully utilize the global geometry information captured by LiDAR and local dynamic motions captured by IMUs, we design a two-stage pose estimator in a coarse-to-fine manner, where point clouds provide the coarse body shape and IMU measurements optimize the local actions. Furthermore, considering the translation deviation caused by the view-dependent partial point cloud, we propose a pose-guided translation corrector. It predicts the offset between captured points and the real root locations, which makes the consecutive movements and trajectories more precise and natural. Moreover, we collect a LiDAR-IMU multi-modal mocap dataset, LIPD, with diverse human actions in long-range scenarios. Extensive quantitative and qualitative experiments on LIPD and other open datasets all demonstrate the capability of our approach for compelling motion capture in large-scale scenarios, which outperforms other methods by an obvious margin. We will release our code and captured dataset to stimulate future research.

Recent neural rendering techniques have greatly benefited image-based modeling and relighting tasks. They provide a continuous, compact, and parallelable representation by modeling the plenoptic function as multilayer perceptrons (MLPs). However, vanilla MLPs suffer from spectral biases on multidimensional datasets. Recent rescues based on isotropic Fourier features mapping mitigate the problem but still fall short of handling heterogeneity across different dimensions, causing imbalanced regression and visual artifacts such as excessive blurs. We present an anisotropic random Fourier features (RFF) mapping scheme to tackle spectral biases. We first analyze the influence of bandwidth from a different perspective: we show that the optimal bandwidth exhibits strong correlations with the frequency spectrum of the training data across various dimensions. We then introduce an anisotropic feature mapping scheme with multiple bandwidths to model the multidimensional signal characteristics. We further propose an efficient bandwidth searching scheme through iterative golden-section search that can significantly reduce the training overload from polynomial time to logarithm. Our anisotropic scheme directly applies to neural surface light-field rendering and image-based relighting. Comprehensive experiments show that our scheme can more faithfully model lighting conditions and object features as well as preserve fine texture details and smooth view transitions even when angular and spatial samples are highly imbalanced.

We propose Human-centered 4D Scene Capture (HSC4D) to accurately and efficiently create a dynamic digital world, containing large-scale indoor-outdoor scenes, diverse human motions, and rich interactions between humans and environments. Using only body-mounted IMUs and LiDAR, HSC4D is space-free without any external devices' constraints and map-free without pre-built maps. Considering that IMUs can capture human poses but always drift for long-period use, while LiDAR is stable for global localization but rough for local positions and orientations, HSC4D makes both sensors complement each other by a joint optimization and achieves promising results for long-term capture. Relationships between humans and environments are also explored to make their interaction more realistic. To facilitate many down-stream tasks, like AR, VR, robots, autonomous driving, etc., we propose a dataset containing three large scenes (1k-5k m2) with accurate dynamic human motions and locations. Diverse scenarios (climbing gym, multi-story building, slope, etc.) and challenging human activities (exercising, walking up/down stairs, climbing, etc.) demonstrate the effectiveness and the generalization ability of HSC4D. The dataset and code is available at lidarhumanmotion.net/hsc4d.

4D modeling of human-object interactions is critical for numerous applications. However, efficient volumetric capture and rendering of complex interaction scenarios, especially from sparse inputs, remain challenging. In this paper, we propose NeuralHOFusion, a neural approach for volumetric human-object capture and rendering using sparse consumer RGBD sensors. It marries traditional non-rigid fusion with recent neural implicit modeling and blending advances, where the captured humans and objects are layer-wise disentangled. For geometry modeling, we propose a neural implicit inference scheme with non-rigid key-volume fusion, as well as a template-aid robust object tracking pipeline. Our scheme enables detailed and complete geometry generation under complex interactions and occlusions. Moreover, we introduce a layer-wise human-object texture rendering scheme, which combines volumetric and image-based rendering in both spatial and temporal domains to obtain photo-realistic results. Extensive experiments demonstrate the effectiveness and efficiency of our approach in synthesizing photo-realistic free-view results under complex human-object interactions.

Existing motion capture datasets are largely short-range and cannot yet fit the need of long-range applications. We propose LiDARHuman26M, a new human motion capture dataset captured by LiDAR at a much longer range to overcome this limitation. Our dataset also includes the ground truth human motions acquired by the IMU system and the synchronous RGB images. We further present a strong base-line method, LiDARCap, for LiDAR point cloud human motion capture. Specifically, we first utilize$PointNet++$to encode features of points and then employ the inverse kinematics solver and SMPL optimizer to regress the pose through aggregating the temporally encoded features hierarchically. Quantitative and qualitative experiments show that our method outperforms the techniques based only on RGB images. Ablation experiments demonstrate that our dataset is challenging and worthy of further research. Finally, the experiments on the KITTI Dataset and the Waymo Open Dataset show that our method can be generalized to different LiDAR sensor settings.

Implicit neural representations such as Neural Radiance Field (NeRF) have focused mainly on modeling static objects captured under multi-view settings where real-time rendering can be achieved with smart data structures, e.g., PlenOctree. In this paper, we present a novel Fourier PlenOctree (FPO) technique to tackle efficient neural mod-eling and real-time rendering of dynamic scenes captured under the free-view video (FVV) setting. The key idea in our FPO is a novel combination of generalized NeRF, PlenOctree representation, volumetric fusion and Fourier transform. To accelerate FPO construction, we present a novel coarse-to-fine fusion scheme that leverages the gen-eralizable NeRF technique to generate the tree via spatial blending. To tackle dynamic scenes, we tailor the implicit network to model the Fourier coefficients of time-varying density and color attributes. Finally, we construct the FPO and train the Fourier coefficients directly on the leaves of a union PlenOctree structure of the dynamic sequence. We show that the resulting FPO enables compact memory overload to handle dynamic objects and supports efficient fine-tuning. Extensive experiments show that the proposed method is 3000 times faster than the original NeRF and achieves over an order of magnitude acceleration over SOTA while preserving high visual quality for the free-viewpoint rendering of unseen dynamic scenes.

Recent neural human representations can produce high-quality multi-view rendering but require using dense multi-view inputs and costly training. They are hence largely limited to static models as training each frame is infeasible. We present HumanNeRF - a neural representation with efficient generalization ability - for high-fidelity free-view synthesis of dynamic humans. Analogous to how IBRNet assists NeRF by avoiding perscene training, HumanNeRF employs an aggregated pixel-alignment feature across multi-view inputs along with a pose embedded non-rigid deformation field for tackling dynamic motions. The raw Human-NeRF can already produce reasonable rendering on sparse video inputs of unseen subjects and camera settings. To further improve the rendering quality, we augment our solution with in-hour scene-specific fine-tuning, and an appearance blending module for combining the benefits of both neural volumetric rendering and neural texture blending. Extensive experiments on various multi-view dynamic hu-man datasets demonstrate effectiveness of our approach in synthesizing photo-realistic free-view humans under challenging motions and with very sparse camera view inputs.

Nearly all existing human pose estimation techniques address the problem under the line-of-sight (LOS) setting. Many real-life applications such as rescue missions and autonomous driving, in contrast, require estimating the pose of hidden subjects. In this paper, we present a non-line-of-sight (NLOS) pose estimator, which produces a skeletal representation of hidden human poses. A brute-force approach would first conduct albedo reconstruction of a hidden subject and then apply LOS pose estimation. We show that such an implementation does not effectively exploit features unique to NLOS and subsequently yields artifacts such as missing joints. We instead first generate a comprehensive NLOS human pose dataset of 19 subjects under 9 motions. We then present a spatially aware deep learning technique based on convolutional neural networks that explicitly employ NLOS features. Comprehensive experiments on both synthetic and real data show that our new estimator is both effective and robust and can be seamlessly integrated into learning-based NLOS scene reconstruction. Our HiddenPose transient dataset contains synthetic transients with ground-truths of the volumes and the joints and real-world transients captured from our NLOS imaging system. Extensive assessments demonstrate that the HiddenPose transient dataset is valuable for effective NLOS research. We will make our data and code publicly available.

Inter-person occlusion and depth ambiguity make estimating the 3D poses of monocular multiple persons as camera-centric coordinates a challenging problem. Typical top-down frameworks suffer from high computational redundancy with an additional detection stage. By contrast, the bottom-up methods enjoy low computational costs as they are less affected by the number of humans. However, most existing bottom-up methods treat camera-centric 3D human pose estimation as two unrelated subtasks: 2.5D pose estimation and camera-centric depth estimation. In this paper, we propose a unified model that leverages the mutual benefits of both these subtasks. Within the framework, a robust structured 2.5D pose estimation is designed to recognize inter-person occlusion based on depth relationships. Additionally, we develop an end-to-end geometry-aware depth reasoning method that exploits the mutual benefits of both 2.5D pose and camera-centric root depths. This method first uses 2.5D pose and geometry information to infer camera-centric root depths in a forward pass, and then exploits the root depths to further improve representation learning of 2.5D pose estimation in a backward pass. Further, we designed an adaptive fusion scheme that leverages both visual perception and body geometry to alleviate inherent depth ambiguity issues. Extensive experiments demonstrate the superiority of our proposed model over a wide range of bottom-up methods. Our accuracy is even competitive with top-down counterparts. Notably, our model runs much faster than existing bottom-up and top-down methods.

Abstract 3D‐aware generative models have demonstrated their superb performance to generate 3D neural radiance fields (NeRF) from a collection of monocular 2D images even for topology‐varying object categories. However, these methods still lack the capability to separately control the shape and appearance of the objects in the generated radiance fields. In this paper, we propose a generative model for synthesizing radiance fields of topology‐varying objects with disentangled shape and appearance variations. Our method generates deformable radiance fields, which builds the dense correspondence between the density fields of the objects and encodes their appearances in a shared template field. Our disentanglement is achieved in an unsupervised manner without introducing extra labels to previous 3D‐aware GAN training. We also develop an effective image inversion scheme for reconstructing the radiance field of an object in a real monocular image and manipulating its shape and appearance. Experiments show that our method can successfully learn the generative model from unstructured monocular images and well disentangle the shape and appearance for objects (e.g., chairs) with large topological variance. The model trained on synthetic data can faithfully reconstruct the real object in a given single image and achieve high‐quality texture and shape editing results.

A densely-sampled light field (LF) is highly desirable in various applications, such as 3-D reconstruction, post-capture refocusing and virtual reality. However, it is costly to acquire such data. Although many computational methods have been proposed to reconstruct a densely-sampled LF from a sparsely-sampled one, they still suffer from either low reconstruction quality, low computational efficiency, or the restriction on the regularity of the sampling pattern. To this end, we propose a novel learning-based method, which accepts sparsely-sampled LFs with irregular structures, and produces densely-sampled LFs with arbitrary angular resolution accurately and efficiently. We also propose a simple yet effective method for optimizing the sampling pattern. Our proposed method, an end-to-end trainable network, reconstructs a densely-sampled LF in a coarse-to-fine manner. Specifically, the coarse sub-aperture image (SAI) synthesis module first explores the scene geometry from an unstructured sparsely-sampled LF and leverages it to independently synthesize novel SAIs, in which a confidence-based blending strategy is proposed to fuse the information from different input SAIs, giving an intermediate densely-sampled LF. Then, the efficient LF refinement module learns the angular relationship within the intermediate result to recover the LF parallax structure. Comprehensive experimental evaluations demonstrate the superiority of our method on both real-world and synthetic LF images when compared with state-of-the-art methods. In addition, we illustrate the benefits and advantages of the proposed approach when applied in various LF-based applications, including image-based rendering and depth estimation enhancement. The code is available at https://github.com/jingjin25/LFASR-FS-GAF.

Light field essentially represents rays in space. The epipolar geometry between two light fields is an important relationship that captures ray-ray correspondences and relative configuration of two views. Unfortunately, so far little work has been done in deriving a formal epipolar geometry model that is specifically tailored for light field cameras. This is primarily due to the high-dimensional nature of the ray sampling process with a light field camera. This paper fills in this gap by developing a novel ray-space epipolar geometry which intrinsically encapsulates the complete projective relationship between two light fields, while the generalized epipolar geometry which describes relationship of normalized light fields is the specialization of the proposed model to calibrated cameras. With Plücker parameterization, we propose the ray-space projection model involving a 6×6 ray-space intrinsic matrix for ray sampling of light field camera. Ray-space fundamental matrix and its properties are then derived to constrain ray-ray correspondences for general and special motions. Finally, based on ray-space epipolar geometry, we present two novel algorithms, one for fundamental matrix estimation, and the other for calibration. Experiments on synthetic and real data have validated the effectiveness of ray-space epipolar geometry in solving 3D computer vision tasks with light field cameras.

A Bayesian nonparametric approach for estimation of a Dirichlet process (DP) mixture of generalized inverted Dirichlet distributions [i.e., an infinite generalized inverted Dirichlet mixture model (InGIDMM)] has been proposed. The generalized inverted Dirichlet distribution has been proven to be efficient in modeling the vectors that contain only positive elements. Under the classical variational inference (VI) framework, the key challenge in the Bayesian estimation of InGIDMM is that the expectation of the joint distribution of data and variables cannot be explicitly calculated. Therefore, numerical methods are usually applied to simulate the optimal posterior distributions. With the recently proposed extended VI (EVI) framework, we introduce lower bound approximations to the original variational objective function in the VI framework such that an analytically tractable solution can be derived. Hence, the problem in numerical simulation has been overcome. By applying the DP mixture technique, an InGIDMM can automatically determine the number of mixture components from the observed data. Moreover, the DP mixture model with an infinite number of mixture components also avoids the problems of underfitting and overfitting. The performance of the proposed approach is demonstrated with both synthesized data and real-life data applications.

Recently, Generative Adversarial Networks (GANs) have been widely used for portrait image generation. However, in the latent space learned by GANs, different attributes, such as pose, shape, and texture style, are generally entangled, making the explicit control of specific attributes difficult. To address this issue, we propose a SofGAN image generator to decouple the latent space of portraits into two subspaces: a geometry space and a texture space. The latent codes sampled from the two subspaces are fed to two network branches separately, one to generate the 3D geometry of portraits with canonical pose, and the other to generate textures. The aligned 3D geometries also come with semantic part segmentation, encoded as a semantic occupancy field (SOF). The SOF allows the rendering of consistent 2D semantic segmentation maps at arbitrary views, which are then fused with the generated texturemaps and stylized to a portrait photo using our semantic instance-wise module. Through extensive experiments, we show that our system can generate high-quality portrait images with independently controllable geometry and texture attributes. The method also generalizes well in various applications, such as appearance-consistent facial animation and dynamic styling.

In this article, we present TightCap, a data-driven scheme to capture both the human shape and dressed garments accurately with only a single three-dimensional (3D) human scan, which enables numerous applications such as virtual try-on, biometrics, and body evaluation. To break the severe variations of the human poses and garments, we propose to model the clothing tightness field—the displacements from the garments to the human shape implicitly in the global UV texturing domain. To this end, we utilize an enhanced statistical human template and an effective multi-stage alignment scheme to map the 3D scan into a hybrid 2D geometry image. Based on this 2D representation, we propose a novel framework to predict clothing tightness field via a novel tightness formulation, as well as an effective optimization scheme to further reconstruct multi-layer human shape and garments under various clothing categories and human postures. We further propose a new clothing tightness dataset of human scans with a large variety of clothing styles, poses, and corresponding ground-truth human shapes to stimulate further research. Extensive experiments demonstrate the effectiveness of our TightCap to achieve the high-quality human shape and dressed garments reconstruction, as well as the further applications for clothing segmentation, retargeting, and animation.

Emerging Metaverse applications demand reliable, accurate, and photorealistic reproductions of human hands to perform sophisticated operations as if in the physical world. While real human hand represents one of the most intricate coordination between bones, muscle, tendon, and skin, state-of-the-art techniques unanimously focus on modeling only the skeleton of the hand. In this paper, we present NIMBLE, a novel parametric hand model that includes the missing key components, bringing 3D hand model to a new level of realism. We first annotate muscles, bones and skins on the recent Magnetic Resonance Imaging hand (MRI-Hand) dataset [Li et al. 2021] and then register a volumetric template hand onto individual poses and subjects within the dataset. NIMBLE consists of 20 bones as triangular meshes, 7 muscle groups as tetrahedral meshes, and a skin mesh. Via iterative shape registration and parameter learning, it further produces shape blend shapes, pose blend shapes, and a joint regressor. We demonstrate applying NIMBLE to modeling, rendering, and visual inference tasks. By enforcing the inner bones and muscles to match anatomic and kinematic rules, NIMBLE can animate 3D hands to new poses at unprecedented realism. To model the appearance of skin, we further construct a photometric HandStage to acquire high-quality textures and normal maps to model wrinkles and palm print. Finally, NIMBLE also benefits learning-based hand pose and shape estimation by either synthesizing rich data or acting directly as a differentiable layer in the inference network.

We, humans, are entering into a virtual era and indeed want to bring animals to the virtual world as well for companion. Yet, computer-generated (CGI) furry animals are limited by tedious off-line rendering, let alone interactive motion control. In this paper, we present ARTEMIS, a novel neural modeling and rendering pipeline for generating ARTiculated neural pets with appEarance and Motion synthesIS. Our ARTEMIS enables interactive motion control, real-time animation, and photo-realistic rendering of furry animals. The core of our ARTEMIS is a neural-generated (NGI) animal engine, which adopts an efficient octree-based representation for animal animation and fur rendering. The animation then becomes equivalent to voxel-level deformation based on explicit skeletal warping. We further use a fast octree indexing and efficient volumetric rendering scheme to generate appearance and density features maps. Finally, we propose a novel shading network to generate high-fidelity details of appearance and opacity under novel poses from appearance and density feature maps. For the motion control module in ARTEMIS, we combine state-of-the-art animal motion capture approach with recent neural character control scheme. We introduce an effective optimization scheme to reconstruct the skeletal motion of real animals captured by a multi-view RGB and Vicon camera array. We feed all the captured motion into a neural character control scheme to generate abstract control signals with motion styles. We further integrate ARTEMIS into existing engines that support VR headsets, providing an unprecedented immersive experience where a user can intimately interact with a variety of virtual animals with vivid movements and photo-realistic appearance. Extensive experiments and showcases demonstrate the effectiveness of our ARTEMIS system in achieving highly realistic rendering of NGI animals in real-time, providing daily immersive and interactive experiences with digital animals unseen before. We make available our ARTEMIS model and dynamic furry animal dataset at https://haiminluo.github.io/publication/artemis/.

Recent years have seen growing interest in 3D human face modeling due to its wide applications in digital human, character generation and animation. Existing approaches overwhelmingly emphasized on modeling the exterior shapes, textures and skin properties of faces, ignoring the inherent correlation between inner skeletal structures and appearance. In this paper, we present SCULPTOR, 3D face creations with Skeleton Consistency Using a Learned Parametric facial generaTOR , aiming to facilitate the easy creation of both anatomically correct and visually convincing face models via a hybrid parametric-physical representation. At the core of SCULPTOR is LUCY, the first large-scale shape-skeleton face dataset in collaboration with plastic surgeons. Named after the fossils of one of the oldest known human ancestors, our LUCY dataset contains high-quality Computed Tomography (CT) scans of the complete human head before and after orthognathic surgeries, which are critical for evaluating surgery results. LUCY consists of 144 scans of 72 subjects (31 male and 41 female), where each subject has two CT scans taken pre- and post-orthognathic operations. Based on our LUCY dataset, we learned a novel skeleton consistent parametric facial generator, SCULPTOR, which can create unique and nuanced facial features that help define a character and at the same time maintain physiological soundness. Our SCULPTOR jointly models the skull, face geometry and face appearance under a unified data-driven framework by separating the depiction of a 3D face into shape blend shape, pose blend shape and facial expression blend shape. SCULPTOR preserves both anatomic correctness and visual realism in facial generation tasks compared with existing methods. Finally, we showcase the robustness and effectiveness of SCULPTOR in various fancy applications unseen before, like archaeological skeletal facial completion, bone-aware character fusion, skull inference from images, face generation with lipo-Level change and facial animations, etc.

The practical deployment of Neural Radiance Fields (NeRF) in rendering applications faces several challenges, with the most critical one being low rendering speed on even high-end graphic processing units (GPUs). In this paper, we present ICARUS, a specialized accelerator architecture tailored for NeRF rendering. Unlike GPUs using general purpose computing and memory architectures for NeRF, ICARUS executes the complete NeRF pipeline using dedicated plenoptic cores (PLCore) consisting of a positional encoding unit (PEU), a multi-layer perceptron (MLP) engine, and a volume rendering unit (VRU). A PLCore takes in positions & directions and renders the corresponding pixel colors without any intermediate data going off-chip for temporary storage and exchange, which can be time and power consuming. To implement the most expensive component of NeRF, i.e., the MLP, we transform the fully connected operations to approximated reconfigurable multiple constant multiplications (MCMs), where common subexpressions are shared across different multiplications to improve the computation efficiency. We build a prototype ICARUS using Synopsys HAPS-80 S104, a field programmable gate array (FPGA)-based prototyping system for large-scale integrated circuits and systems design. We evaluate the power-performancearea (PPA) of a PLCore using 40nm LP CMOS technology. Working at 400 MHz, a single PLCore occupies 16.5 mm 2 and consumes 282.8 mW, translating to 0.105 uJ/sample. The results are compared with those of GPU and tensor processing unit (TPU) implementations.

Production-level workflows for producing convincing 3D dynamic human faces have long relied on an assortment of labor-intensive tools for geometry and texture generation, motion capture and rigging, and expression synthesis. Recent neural approaches automate individual components but the corresponding latent representations cannot provide artists with explicit controls as in conventional tools. In this paper, we present a new learning-based, video-driven approach for generating dynamic facial geometries with high-quality physically-based assets. For data collection, we construct a hybrid multiview-photometric capture stage, coupling with ultra-fast video cameras to obtain raw 3D facial assets. We then set out to model the facial expression, geometry and physically-based textures using separate VAEs where we impose a global MLP based expression mapping across the latent spaces of respective networks, to preserve characteristics across respective attributes. We also model the delta information as wrinkle maps for the physically-based textures, achieving high-quality 4K dynamic textures. We demonstrate our approach in high-fidelity performer-specific facial capture and cross-identity facial motion retargeting. In addition, our multi-VAE-based neural asset, along with the fast adaptation schemes, can also be deployed to handle in-the-wild videos. Besides, we motivate the utility of our explicit facial disentangling strategy by providing various promising physically-based editing results with high realism. Comprehensive experiments show that our technique provides higher accuracy and visual fidelity than previous video-driven facial reconstruction and animation methods.

We have recently seen tremendous progress in the neural advances for photo-real human modeling and rendering. However, it's still challenging to integrate them into an existing mesh-based pipeline for downstream applications. In this paper, we present a comprehensive neural approach for high-quality reconstruction, compression, and rendering of human performances from dense multi-view videos. Our core intuition is to bridge the traditional animated mesh workflow with a new class of highly efficient neural techniques. We first introduce a neural surface reconstructor for high-quality surface generation in minutes. It marries the implicit volumetric rendering of the truncated signed distance field (TSDF) with multi-resolution hash encoding. We further propose a hybrid neural tracker to generate animated meshes, which combines explicit non-rigid tracking with implicit dynamic deformation in a self-supervised framework. The former provides the coarse warping back into the canonical space, while the latter implicit one further predicts the displacements using the 4D hash encoding as in our reconstructor. Then, we discuss the rendering schemes using the obtained animated meshes, ranging from dynamic texturing to lumigraph rendering under various bandwidth settings. To strike an intricate balance between quality and bandwidth, we propose a hierarchical solution by first rendering 6 virtual views covering the performer and then conducting occlusion-aware neural texture blending. We demonstrate the efficacy of our approach in a variety of mesh-based applications and photo-realistic free-view experiences on various platforms, i.e., inserting virtual human performances into real environments through mobile AR or immersively watching talent shows with VR headsets.

Photo-realistic modeling and rendering of fuzzy objects with complex opacity are critical for numerous immersive VR/AR applications, but it suffers from strong view-dependent brightness, color. In this paper, we propose a novel scheme to generate opacity radiance fields with a convolutional neural renderer for fuzzy objects, which is the first to combine both explicit opacity supervision and convolutional mechanism into the neural radiance field framework so as to enable high-quality appearance and global consistent alpha mattes generation in arbitrary novel views. More specifically, we propose an efficient sampling strategy along with both the camera rays and image plane, which enables efficient radiance field sampling and learning in a patch-wise manner, as well as a novel volumetric feature integration scheme that generates per-patch hybrid feature embeddings to reconstruct the view-consistent fine-detailed appearance and opacity output. We further adopt a patch-wise adversarial training scheme to preserve both high-frequency appearance and opacity details in a self-supervised framework. We also introduce an effective multi-view image capture system to capture high-quality color and alpha maps for challenging fuzzy objects. Extensive experiments on existing and our new challenging fuzzy object dataset demonstrate that our method achieves photo-realistic, globally consistent, and fined detailed appearance and opacity free-viewpoint rendering for various fuzzy objects.

Photo-realistic neural reconstruction and rendering of the human portrait are critical for numerous VR/AR applications. Still, existing solutions inherently rely on multi-view capture settings, and the one-shot solution to get rid of the tedious multi-view synchronization and calibration remains extremely challenging. In this paper, we propose MirrorNeRF - a one-shot neural portrait free-viewpoint rendering approach using a catadioptric imaging system with multiple sphere mirrors and a single high-resolution digital camera, which is the first to combine neural radiance field with catadioptric imaging so as to enable one-shot photo-realistic human portrait reconstruction and rendering, in a low-cost and casual capture setting. More specifically, we propose a light-weight catadioptric system design with a sphere mirror array to enable diverse ray sampling in the continuous 3D space as well as an effective online calibration for the camera and the mirror array. Our catadioptric imaging system can be easily deployed with a low budget and the casual capture ability for convenient daily usages. We introduce a novel neural warping radiance field representation to learn a continuous displacement field that implicitly compensates for the misalignment due to our flexible system setting. We further propose a density regularization scheme to leverage the inherent geometry information from the catadioptric data in a self-supervision manner, which not only improves the training efficiency but also provides more effective density supervision for higher rendering quality. Extensive experiments demonstrate the effectiveness and robustness of our scheme to achieve one-shot photo-realistic and high-quality appearance free-viewpoint rendering for human portrait scenes.

We present MVSNeRF, a novel neural rendering approach that can efficiently reconstruct neural radiance fields for view synthesis. Unlike prior works on neural radiance fields that consider per-scene optimization on densely captured images, we propose a generic deep neural network that can reconstruct radiance fields from only three nearby input views via fast network inference. Our approach leverages plane-swept cost volumes (widely used in multi-view stereo) for geometry-aware scene reasoning, and combines this with physically based volume rendering for neural radiance field reconstruction. We train our network on real objects in the DTU dataset, and test it on three different datasets to evaluate its effectiveness and generalizability. Our approach can generalize across scenes (even indoor scenes, completely different from our training scenes of objects) and generate realistic view synthesis results using only three input images, significantly outperforming concurrent works on generalizable radiance field reconstruction. Moreover, if dense images are captured, our estimated radiance field representation can be easily fine-tuned; this leads to fast per-scene reconstruction with higher rendering quality and substantially less optimization time than NeRF.

We introduce GNeRF, a framework to marry Generative Adversarial Networks (GAN) with Neural Radiance Field (NeRF) reconstruction for the complex scenarios with unknown and even randomly initialized camera poses. Recent NeRF-based advances have gained popularity for remarkable realistic novel view synthesis. However, most of them heavily rely on accurate camera poses estimation, while few recent methods can only optimize the unknown camera poses in roughly forward-facing scenes with relatively short camera trajectories and require rough camera poses initialization. Differently, our GNeRF only utilizes randomly initialized poses for complex outside-in scenarios. We propose a novel two-phases end-to-end framework. The first phase takes the use of GANs into the new realm for optimizing coarse camera poses and radiance fields jointly, while the second phase refines them with additional photometric loss. We overcome local minima using a hybrid and iterative optimization scheme. Extensive experiments on a variety of synthetic and natural scenes demonstrate the effectiveness of GNeRF. More impressively, our approach outperforms the baselines favorably in those scenes with repeated patterns or even low textures that are regarded as extremely challenging before.

Video portraits relighting is critical in user-facing human photography, especially for immersive VR/AR experience. Recent advances still fail to recover consistent relit result under dynamic illuminations from monocular RGB stream, suffering from the lack of video consistency supervision. In this paper, we propose a neural approach for real-time, high-quality and coherent video portrait relighting, which jointly models the semantic, temporal and lighting consistency using a new dynamic OLAT dataset. We propose a hybrid structure and lighting disentanglement in an encoder-decoder architecture, which combines a multi-task and adversarial training strategy for semantic-aware consistency modeling. We adopt a temporal modeling scheme via flow-based supervision to encode the conjugated temporal consistency in a cross manner. We also propose a lighting sampling strategy to model the illumination consistency and mutation for natural portrait light manipulation in real-world. Extensive experiments demonstrate the effectiveness of our approach for consistent video portrait light-editing and relighting, even using mobile computing.

Hand modeling is critical for immersive VR/AR, action understanding, or human healthcare. Existing parametric models account only for hand shape, pose, or texture, without modeling the anatomical attributes like bone, which is essential for realistic hand biomechanics analysis. In this paper, we present PIANO, the first parametric bone model of human hands from MRI data. Our PIANO model is biologically correct, simple to animate, and differentiable, achieving more anatomically precise modeling of the inner hand kinematic structure in a data-driven manner than the traditional hand models based on the outer surface only. Furthermore, our PIANO model can be applied in neural network layers to enable training with a fine-grained semantic loss, which opens up the new task of data-driven fine-grained hand bone anatomic and semantic understanding from MRI or even RGB images. We make our model publicly available.

Recent neural rendering approaches for human activities achieve remarkable view synthesis results, but still rely on dense input views or dense training with all the capture frames, leading to deployment difficulty and inefficient training overload. However, existing advances will be ill-posed if the input is both spatially and temporally sparse. To fill this gap, in this paper we propose a few-shot neural human rendering approach (FNHR) from only sparse RGBD inputs, which exploits the temporal and spatial redundancy to generate photo-realistic free-view output of human activities. Our FNHR is trained only on the key-frames which expand the motion manifold in the input sequences. We introduce a two-branch neural blending to combine the neural point render and classical graphics texturing pipeline, which integrates reliable observations over sparse key-frames. Furthermore, we adopt a patch-based adversarial training process to make use of the local redundancy and avoids over-fitting to the key-frames, which generates fine-detailed rendering results. Extensive experiments demonstrate the effectiveness of our approach to generate high-quality free view-point results for challenging human performances under the sparse setting.

4D reconstruction of human-object interaction is critical for immersive VR/AR experience and human activity understanding. Recent advances still fail to recover fine geometry and texture results from sparse RGB inputs, especially under challenging human-object interactions scenarios. In this paper, we propose a neural human performance capture and rendering system to generate both high-quality geometry and photo-realistic texture of both human and objects under challenging interaction scenarios in arbitrary novel views, from only sparse RGB streams. To deal with complex occlusions raised by human-object interactions, we adopt a layer-wise scene decoupling strategy and perform volumetric reconstruction and neural rendering of the human and object. Specifically, for geometry reconstruction, we propose an interaction-aware human-object capture scheme that jointly considers the human reconstruction and object reconstruction with their correlations. Occlusion-aware human reconstruction and robust human-aware object tracking are proposed for consistent 4D human-object dynamic reconstruction. For neural texture rendering, we propose a layer-wise human-object rendering scheme, which combines direction-aware neural blending weight learning and spatial-temporal texture completion to provide high-resolution and photo-realistic texture results in the occluded scenarios. Extensive experiments demonstrate the effectiveness of our approach to achieve high-quality geometry and texture reconstruction in free viewpoints for challenging human-object interactions.

Generating "bullet-time" effects of human free-viewpoint videos is critical for immersive visual effects and VR/AR experience. Recent neural advances still lack the controllable and interactive bullet-time design ability for human free-viewpoint rendering, especially under the real-time, dynamic and general setting for our trajectory-aware task. To fill this gap, in this paper we propose a neural interactive bullet-time generator (iButter) for photo-realistic human free-viewpoint rendering from dense RGB streams, which enables flexible and interactive design for human bullet-time visual effects. Our iButter approach consists of a real-time preview and design stage as well as a trajectory-aware refinement stage. During preview, we propose an interactive bullet-time design approach by extending the NeRF rendering to a real-time and dynamic setting and getting rid of the tedious per-scene training. To this end, our bullet-time design stage utilizes a hybrid training set, light-weight network design and an efficient silhouette-based sampling strategy. During refinement, we introduce an efficient trajectory-aware scheme within 20 minutes, which jointly encodes the spatial, temporal consistency and semantic cues along the designed trajectory, achieving photo-realistic bullet-time viewing experience of human activities. Extensive experiments demonstrate the effectiveness of our approach for convenient interactive bullet-time design and photo-realistic human free-viewpoint video generation.

We present a neural modeling framework for non-line-of-sight (NLOS) imaging. Previous solutions have sought to explicitly recover the 3D geometry (e.g., as point clouds) or voxel density (e.g., within a pre-defined volume) of the hidden scene. In contrast, inspired by the recent Neural Radiance Field (NeRF) approach, we use a multi-layer perceptron (MLP) to represent the neural transient field or NeTF. However, NeTF measures the transient over spherical wavefronts rather than the radiance along lines. We therefore formulate a spherical volume NeTF reconstruction pipeline, applicable to both confocal and non-confocal setups. Compared with NeRF, NeTF samples a much sparser set of viewpoints (scanning spots) and the sampling is highly uneven. We thus introduce a Monte Carlo technique to improve the robustness in the reconstruction. Experiments on synthetic and real datasets demonstrate NeTF achieves state-of-the-art performance and can provide reliable reconstructions even under semi-occlusions and on non-Lambertian materials.

We present a structure-from-motion (SfM) framework based on a special type of multi-perspective camera called the cross-slit or XSlit camera. Traditional perspective camera based SfM suffers from the scale ambiguity which is inherent to the pinhole camera geometry. In contrast, an XSlit camera captures rays passing through two oblique lines in 3D space and we show such ray geometry directly resolves the scale ambiguity when employed for SfM. To accommodate the XSlit cameras, we develop tailored feature matching, camera pose estimation, triangulation, and bundle adjustment techniques. Specifically, we devise a SIFT feature variant using non-uniform Gaussian kernels to handle the distortions in XSlit images for reliable feature matching. Moreover, we demonstrate that the XSlit camera exhibits ambiguities in pose estimation process which can not be handled by existing work. Consequently, we propose a 14 point algorithm to properly handle the XSlit degeneracy and estimate the relative pose between XSlit cameras from feature correspondences. We further exploit the unique depth-dependent aspect ratio (DDAR) property to improve the bundle adjustment for the XSlit camera. Synthetic and real experiments demonstrate that the proposed XSlit SfM can conduct reliable and high fidelity 3D reconstruction at an absolute scale.

Generating free-viewpoint videos is critical for immersive VR/AR experience, but recent neural advances still lack the editing ability to manipulate the visual perception for large dynamic scenes. To fill this gap, in this paper, we propose the first approach for editable free-viewpoint video generation for large-scale view-dependent dynamic scenes using only 16 cameras. The core of our approach is a new layered neural representation, where each dynamic entity, including the environment itself, is formulated into a spatio-temporal coherent neural layered radiance representation called ST-NeRF. Such a layered representation supports manipulations of the dynamic scene while still supporting a wide free viewing experience. In our ST-NeRF, we represent the dynamic entity/layer as a continuous function, which achieves the disentanglement of location, deformation as well as the appearance of the dynamic entity in a continuous and self-supervised manner. We propose a scene parsing 4D label map tracking to disentangle the spatial information explicitly and a continuous deform module to disentangle the temporal motion implicitly. An object-aware volume rendering scheme is further introduced for the re-assembling of all the neural layers. We adopt a novel layered loss and motion-aware ray sampling strategy to enable efficient training for a large dynamic scene with multiple performers, Our framework further enables a variety of editing functions, i.e., manipulating the scale and location, duplicating or retiming individual neural layers to create numerous visual effects while preserving high realism. Extensive experiments demonstrate the effectiveness of our approach to achieve high-quality, photo-realistic, and editable free-viewpoint video generation for dynamic scenes.

Efficient and accurate segmentation of full 4D light fields is an important task in computer vision and computer graphics. The massive volume and the redundancy of light fields make it an open challenge. In this article, we propose a novel light field hypergraph (LFHG) representation using the light field super-pixel (LFSP) for interactive light field segmentation. The LFSPs not only maintain the light field spatio-angular consistency, but also greatly contribute to the hypergraph coarsening. These advantages make LFSPs useful to improve segmentation performance. Based on the LFHG representation, we present an efficient light field segmentation algorithm via graph-cut optimization. Experimental results on both synthetic and real scene data demonstrate that our method outperforms state-of-the-art methods on the light field segmentation task with respect to both accuracy and efficiency.

There have been significant advances in capturing gigapixel panoramas (GPP). However, solutions for viewing GPPs on head-mounted displays (HMDs) are lagging: an immersive experience requires ultra-fast rendering while directly loading a GPP onto the GPU is infeasible due to limited texture memory capacity. In this paper, we present a novel out-of-core rendering technique that supports not only classic panning, tilting, and zooming but also dynamic refocusing for viewing a GPP on HMD. Inspired by the network package transmission mechanisms in distributed visualization, our approach employs hierarchical image tiling and on-demand data updates across the main and the GPU memory. We further present a multi-resolution rendering scheme and a refocused light field rendering technique based on RGBD GPPs with minimal memory overhead. Comprehensive experiments demonstrate that our technique is highly efficient and reliable, able to achieve ultra-high frame rates ( fps) even on low-end GPUs. With an embedded gaze tracker, our technique enables immersive panorama viewing experiences with unprecedented resolutions, field-of-view, and focus variations while maintaining smooth spatial, angular, and focal transitions.

We present a novel Relightable Neural Renderer (RNR) for simultaneous view synthesis and relighting using multi-view image inputs. Existing neural rendering (NR) does not explicitly model the physical rendering process and hence has limited capabilities on relighting. RNR instead models image formation in terms of environment lighting, object intrinsic attributes, and light transport function (LTF), each corresponding to a learnable component. In particular, the incorporation of a physically based rendering process not only enables relighting but also improves the quality of view synthesis. Comprehensive experiments on synthetic and real data show that RNR provides a practical and effective solution for conducting free-viewpoint relighting.

Motivated by the correlation between the depth and the geometric structure of a 360 indoor image, we propose a novel learning-based depth estimation framework that leverages the geometric structure of a scene to conduct depth estimation. Specifically, we represent the geometric structure of an indoor scene as a collection of corners, boundaries and planes. On the one hand, once a depth map is estimated, this geometric structure can be inferred from the estimated depth map; thus, the geometric structure functions as a regularizer for depth estimation. On the other hand, this estimation also benefits from the geometric structure of a scene estimated from an image where the structure functions as a prior. However, furniture in indoor scenes makes it challenging to infer geometric structure from depth or image data. An attention map is inferred to facilitate both depth estimation from features of the geometric structure and also geometric inferences from the estimated depth map. To validate the effectiveness of each component in our framework under controlled conditions, we render a synthetic dataset, Shanghaitech-Kujiale Indoor 360 dataset with 3550 360 indoor images. Extensive experiments on popular datasets validate the effectiveness of our solution. We also demonstrate that our method can also be applied to counterfactual depth.

We present an end-to-end Neural Human Renderer (NHR) for dynamic human captures under the multi-view setting. NHR adopts PointNet++ for feature extraction (FE) to enable robust 3D correspondence matching on low quality, dynamic 3D reconstructions. To render new views, we map 3D features onto the target camera as a 2D feature map and employ an anti-aliased CNN to handle holes and noises. Newly synthesized views from NHR can be further used to construct visual hulls to handle textureless and/or dark regions such as black clothing. Comprehensive experiments show NHR significantly outperforms the state-of-the-art neural and image-based rendering techniques, especially on hands, hair, nose, foot, etc.

We present a new color photometric stereo (CPS) method that recovers high quality, detailed 3D face geometry in a single shot. Our system uses three uncalibrated near point lights of different colors and a single camera. For robust self-calibration of the light sources, we use 3D morphable model (3DMM) [1] and semantic segmentation of facial parts. For reconstruction, we address the inherent spectral ambiguity in color photometric stereo by incorporating albedo consensus, albedo similarity, and proxy prior into a unified framework. In this way, we jointly exploit multiple cues to resolve under-determinedness, without the need for spatial constancy of albedo. Experiments show that our new approach produces state-of-the-art results from single image with high-fidelity geometry that includes details such as wrinkles.

This paper explores the problem of reconstructing high-resolution light field (LF) images from hybrid lenses, including a high-resolution camera surrounded by multiple low-resolution cameras. To tackle this challenge, we propose a novel end-to-end learning-based approach, which can comprehensively utilize the specific characteristics of the input from two complementary and parallel perspectives. Specifically, one module regresses a spatially consistent intermediate estimation by learning a deep multidimensional and cross-domain feature representation; the other one constructs another intermediate estimation, which maintains the high-frequency textures, by propagating the information of the high-resolution view. We finally leverage the advantages of the two intermediate estimations via the learned attention maps, leading to the final high-resolution LF image. Extensive experiments demonstrate the significant superiority of our approach over state-of-the-art ones. That is, our method not only improves the PSNR by more than 2 dB, but also preserves the LF structure much better. To the best of our knowledge, this is the first end-to-end deep learning method for reconstructing a high-resolution LF image with a hybrid input. We believe our framework could potentially decrease the cost of high-resolution LF data acquisition and also be beneficial to LF data storage and transmission. The code is available at https://github.com/jingjin25/LFhybridSR-Fusion.

We present a novel real-time line segment detection scheme called Line Graph Neural Network (LGNN). Existing approaches require a computationally expensive verification or postprocessing step. Our LGNN employs a deep convolutional neural network (DCNN) for proposing line segment directly, with a graph neural network (GNN) module for reasoning their connectivities. Specifically, LGNN exploits a new quadruplet representation for each segment where the GNN module takes the predicted candidates as vertexes and constructs a sparse graph to enforce structural context. Compared with the state-of-the-art, LGNN achieves near real-time performance without compromising accuracy. LGNN further enables time-sensitive 3D applications. When a 3D point cloud is accessible, we present a multi-modal line segment classification technique for extracting a 3D wireframe of the environment robustly and efficiently.

Reconstructing a human portrait in a realistic and convenient manner is critical for human modeling and understanding. Aiming at light-weight and realistic human portrait reconstruction, in this paper we propose Neural3D: a novel neural human portrait scanning system using only a single RGB camera. In our system, to enable accurate pose estimation,we propose a context-aware correspondence learning approach which jointly models the appearance, spatial and motion information between feature pairs. To enable realistic reconstruction and suppress the geometry error, we further adopt a point-based neural rendering scheme to generate realistic and immersive portrait visualization in arbitrary virtual view-points. By introducing these learning-based technical components into the pure RGB-based human modeling framework, we can achieve both accurate camera pose estimation and realistic free-viewpoint rendering of the reconstructed human portrait. Extensive experiments on a variety of challenging capture scenarios demonstrate the robustness and effectiveness of our approach.

Fuzzy objects composed of hair, fur, or feather are impossible to scan even with the latest active or passive 3D scanners. We present a novel and practical neural rendering (NR) technique called neural opacity point cloud (NOPC) to allow high quality rendering of such fuzzy objects at any viewpoint. NOPC employs a learning-based scheme to extract geometric and appearance features on 3D point clouds including their opacity. It then maps the 3D features onto virtual viewpoints where a new U-Net based NR manages to handle noisy and incomplete geometry while maintaining translation equivariance. Comprehensive experiments on existing and new datasets show our NOPC can produce photorealistic rendering on inputs from multi-view setups such as a turntable system for hair and furry toy captures.

Recovering the shape and reflectance of non-Lambertian surfaces remains a challenging problem in computer vision since the view-dependent appearance invalidates traditional photo-consistency constraint. In this paper, we introduce a novel concentric multi-spectral light field (CMSLF) design that is able to recover the shape and reflectance of surfaces of various materials in one shot. Our CMSLF system consists of an array of cameras arranged on concentric circles where each ring captures a specific spectrum. Coupled with a multi-spectral ring light, we are able to sample viewpoint and lighting variations in a single shot via spectral multiplexing. We further show that our concentric camera and light source setting results in a unique single-peak pattern in specularity variations across viewpoints. This property enables robust depth estimation for specular points. To estimate depth and multi-spectral reflectance map, we formulate a physics-based reflectance model for the CMSLF under the surface camera (S-Cam) representation. Extensive synthetic and real experiments show that our method outperforms the state-of-the-art shape reconstruction methods, especially for non-Lambertian surfaces.

A deep neural network of multiple nonlinear layers forms a large function space, which can easily lead to overfitting when it encounters small-sample data. To mitigate overfitting in small-sample classification, learning more discriminative features from small-sample data is becoming a new trend. To this end, this paper aims to find a subspace of neural networks that can facilitate a large decision margin. Specifically, we propose the Orthogonal Softmax Layer (OSL), which makes the weight vectors in the classification layer remain orthogonal during both the training and test processes. The Rademacher complexity of a network using the OSL is only 1/K, where K is the number of classes, of that of a network using the fully connected classification layer, leading to a tighter generalization error bound. Experimental results demonstrate that the proposed OSL has better performance than the methods used for comparison on four small-sample benchmark datasets, as well as its applicability to large-sample datasets. Codes are available at: https://github.com/dongliangchang/OSLNet.

Incorrect saliency detection such as false alarms and missed alarms may lead to potentially severe consequences in various application areas. Effective separation of salient objects in complex scenes is a major challenge in saliency detection. In this paper, we propose a new method for saliency detection on light field to improve the saliency detection in challenging scenes. We construct an object-guided depth map, which acts as an inducer to efficiently incorporate the relations among light field cues, by using abundant light field cues. Furthermore, we enforce spatial consistency by constructing an optimization model, named Depth-induced Cellular Automata (DCA), in which the saliency value of each superpixel is updated by exploiting the intrinsic relevance of its similar regions. Additionally, the proposed DCA model enables inaccurate saliency maps to achieve a high level of accuracy. We analyze our approach on one publicly available dataset. Experiments show the proposed method is robust to a wide range of challenging scenes and outperforms the state-of-the-art 2D/3D/4D (light-field) saliency detection approaches.

Recent progresses in visual tracking have greatly improved the tracking performance. However, challenges such as occlusion and view change remain obstacles in real world deployment. A natural solution to these challenges is to use multiple cameras with multiview inputs, though existing systems are mostly limited to specific targets (e.g. human), static cameras, and/or require camera calibration. To break through these limitations, we propose a generic multiview tracking (GMT) framework that allows camera movement, while requiring neither specific object model nor camera calibration. A key innovation in our framework is a cross-camera trajectory prediction network (TPN), which implicitly and dynamically encodes camera geometric relations, and hence addresses missing target issues such as occlusion. Moreover, during tracking, we assemble information across different cameras to dynamically update a novel collaborative correlation filter (CCF), which is shared among cameras to achieve robustness against view change. The two components are integrated into a correlation filter tracking framework, where features are trained offline using existing single view tracking datasets. For evaluation, we first contribute a new generic multiview tracking dataset (GMTD) with careful annotations, and then run experiments on the GMTD and CAMPUS datasets. The proposed GMT algorithm shows clear advantages in terms of robustness over state-of-the-art ones.

This paper presents a fully automatic framework for extracting editable 3D objects directly from a single photograph. Unlike previous methods which recover either depth maps, point clouds, or mesh surfaces, we aim to recover 3D objects with semantic parts and can be directly edited. We base our work on the assumption that most human-made objects are constituted by parts and these parts can be well represented by generalized primitives. Our work makes an attempt towards recovering two types of primitive-shaped objects, namely, generalized cuboids and generalized cylinders. To this end, we build a novel instance-aware segmentation network for accurate part separation. Our GeoNet outputs a set of smooth part-level masks labeled as profiles and bodies. Then in a key stage, we simultaneously identify profile-body relations and recover 3D parts by sweeping the recognized profile along their body contour and jointly optimize the geometry to align with the recovered masks. Qualitative and quantitative experiments show that our algorithm can recover high quality 3D models and outperforms existing methods in both instance segmentation and 3D reconstruction.

This paper tackles RGBD based gaze estimation with Convolutional Neural Networks (CNNs). Specifically, we propose to decompose gaze point estimation into eyeball pose, head pose, and 3D eye position estimation. Compared with RGB image-based gaze tracking, having depth modality helps to facilitate head pose estimation and 3D eye position estimation. The captured depth image, however, usually contains noise and black holes which noticeably hamper gaze tracking. Thus we propose a CNN-based multi-task learning framework to simultaneously refine depth images and predict gaze points. We utilize a generator network for depth image generation with a Generative Neural Network (GAN), where the generator network is partially shared by both the gaze tracking network and GAN-based depth synthesizing. By optimizing the whole network simultaneously, depth image synthesis improves gaze point estimation and vice versa. Since the only existing RGBD dataset (EYEDIAP) is too small, we build a large-scale RGBD gaze tracking dataset for performance evaluation. As far as we know, it is the largest RGBD gaze dataset in terms of the number of participants. Comprehensive experiments demonstrate that our method outperforms existing methods by a large margin on both our dataset and the EYEDIAP dataset.

Light field essentially represents the collection of rays in space. The rays captured by multiple light field cameras form subsets of full rays in 3D space and can be transformed to each other. However, most previous approaches model the projection from an arbitrary point in 3D space to corresponding pixel on the sensor. There are few models on describing the ray sampling and transformation among multiple light field cameras. In the paper, we propose a novel ray-space projection model to transform sets of rays captured by multiple light field cameras in term of the Plucker coordinates. We first derive a 6×6 ray-space intrinsic matrix based on multi-projection-center (MPC) model. A homogeneous ray-space projection matrix and a fundamental matrix are then proposed to establish ray-ray correspondences among multiple light fields. Finally, based on the ray-space projection matrix, a novel camera calibration method is proposed to verify the proposed model. A linear constraint and a ray-ray cost function are established for linear initial solution and non-linear optimization respectively. Experimental results on both synthetic and real light field data have verified the effectiveness and robustness of the proposed model.

Mirror surfaces are notoriously difficult to reconstruct. In this paper, we present a novel computational imaging approach for reconstructing complex mirror surfaces using a dense illumination field with angularly varying polarization states, which we call the polarization field. Specifically, we generate the polarization field using a commercial LCD with the top polarizer removed. We mathematically model the liquid crystals as polarization rotators using Jones calculus and show that the rotated polarization states of outgoing rays encode angular information (e.g., ray directions). To model reflection under the polarization field, we derive a reflection image formation model based on the Fresnel's equations and estimate incident ray positions and directions by coding the polarization field. Finally, we triangulate the incident rays with the camera rays to recover normals/depths of the mirror surface. Comprehensive simulations and real experiments demonstrate the effectiveness of our approach.

We present a single-image 3D face synthesis technique that can handle challenging facial expressions while recovering fine geometric details. Our technique employs expression analysis for proxy face geometry generation and combines supervised and unsupervised learning for facial detail synthesis. On proxy generation, we conduct emotion prediction to determine a new expression-informed proxy. On detail synthesis, we present a Deep Facial Detail Net (DFDN) based on Conditional Generative Adversarial Net (CGAN) that employs both geometry and appearance loss functions. For geometry, we capture 366 high-quality 3D scans from 122 different subjects under 3 facial expressions. For appearance, we use additional 163K in-the-wild face images and apply image-based rendering to accommodate lighting variations. Comprehensive experiments demonstrate that our framework can produce high-quality 3D faces with realistic details under challenging facial expressions.

Reliable human parts segmentation on 2D images plays an important role in many human-centric computer vision tasks. While significant achievements have been made on human pose estimation, the performance on human parts segmentation remains low. In this paper, we present a novel technique that we call Pose2Body that robustly conducts human parts segmentation based on the pose estimation results. We partition an image into superpixels and set out to assign a segment label to each superpixel most consistent with the pose. We design special feature vectors for every superpixel-label assignment as well as superpixel-superpixel pairs and model optimal labeling as to solve for a conditional random field (CRF). Comprehensive experiments show that our technique achieves substantial improvements over the state-of-the-art solutions.

In this work, we demonstrate yet another approach to tackle the amodal segmentation problem. Specifically, we first introduce a new representation, namely a semantics-aware distance map (sem-dist map), to serve as our target for amodal segmentation instead of the commonly used masks and heatmaps. The sem-dist map is a kind of level-set representation, of which the different regions of an object are placed into different levels on the map according to their visibility. It is a natural extension of masks and heatmaps, where modal, amodal segmentation, as well as depth order information, are all well-described. Then we also introduce a novel convolutional neural network (CNN) architecture, which we refer to as semantic layering network, to estimate sem-dist maps layer by layer, from the global-level to the instance-level, for all objects in an image. Extensive experiments on the COCOA and D2SA datasets have demonstrated that our framework can predict amodal segmentation, occlusion, and depth order with state-of-the-art performance.

Camera Response Functions (CRFs) map the irradiance incident at a sensor pixel to an intensity value in the corresponding image pixel. The nonlinearity of CRFs impact physics-based and low-level computer vision methods like de-blurring, photometric stereo, etc. In addition, CRFs have been used for forensics to identify regions of an image spliced in from a different camera. Despite its importance, the process of radiometrically calibrating a camera's CRF is significantly harder and less standardized than geometric calibration. Competing methods use different mathematical models of the CRF, some of which are derived from an outdated dataset. We present a new dataset of 178 CRFs from modern digital cameras, derived from 1565 camera review images available online, and use it to answer a series of questions about CRFs. Which mathematical models are best for CRF estimation? How have they changed over time? And how unique are CRFs from camera to camera?

Nearly all existing visual saliency models by far have focused on predicting a universal saliency map across all observers. Yet psychology studies suggest that visual attention of different observers can vary significantly under specific circumstances, especially a scene is composed of multiple salient objects. To study such heterogenous visual attention pattern across observers, we first construct a personalized saliency dataset and explore correlations between visual attention, personal preferences, and image contents. Specifically, we propose to decompose a personalized saliency map (referred to as PSM) into a universal saliency map (referred to as USM) predictable by existing saliency detection models and a new discrepancy map across users that characterizes personalized saliency. We then present two solutions towards predicting such discrepancy maps, i.e., a multi-task convolutional neural network (CNN) framework and an extended CNN with Person-specific Information Encoded Filters (CNN-PIEF). Extensive experimental results demonstrate the effectiveness of our models for PSM prediction as well their generalization capability for unseen observers.

The goal of image pre-compensation is to process an image such that after being convolved with a known kernel, will appear close to the sharp reference image. In a practical setting, the pre-compensated image has significantly higher dynamic range than the latent image. As a result, some form of tone mapping is needed. In this paper, we show how global tone mapping functions affect contrast and ringing in image pre-compensation. We further enhance contrast and reduce ringing by considering the visual saliency. Specifically, we prioritize contrast preservation in salient regions while tolerating more blurriness elsewhere. For quantitative analysis, we design new metrics to measure the contrast of an image with ringing. Specifically, we set out to find its "equivalent ringing-free" image that matches its intensity histogram and uses its contrast as the measure. We illustrate our approach on projector defocus compensation and visual acuity enhancement. Compared with the state-of-the-art, our approach significantly improves the contrast. We also perform user studies to demonstrate that our method can effectively improve the viewing experience for users with impaired vision.

Light field cameras can capture both spatial and angular information of light rays, enabling 3D reconstruction by a single exposure. The geometry of 3D reconstruction is affected by intrinsic parameters of a light field camera significantly. In the paper, we propose a multi-projection-center (MPC) model with 6 intrinsic parameters to characterize light field cameras based on traditional two-parallel-plane (TPP) representation. The MPC model can generally parameterize light field in different imaging formations, including conventional and focused light field cameras. By the constraints of 4D ray and 3D geometry, a 3D projective transformation is deduced to describe the relationship between geometric structure and the MPC coordinates. Based on the MPC model and projective transformation, we propose a calibration algorithm to verify our light field camera model. Our calibration method includes a close-form solution and a non-linear optimization by minimizing re-projection errors. Experimental results on both simulated and real scene data have verified the performance of our algorithm.

In this paper, we describe how scene depth can be extracted using a hyperspectral light field capture (H-LF) system. Our H-LF system consists of a 5 ×6 array of cameras, with each camera sampling a different narrow band in the visible spectrum. There are two parts to extracting scene depth. The first part is our novel cross-spectral pairwise matching technique, which involves a new spectral-invariant feature descriptor and its companion matching metric we call bidirectional weighted normalized cross correlation (BWNCC). The second part, namely, H-LF stereo matching, uses a combination of spectral-dependent correspondence and defocus cues. These two new cost terms are integrated into a Markov Random Field (MRF) for disparity estimation. Experiments on synthetic and real H-LF data show that our approach can produce high-quality disparity maps. We also show that these results can be used to produce the complete plenoptic cube in addition to synthesizing all-focus and defocused color images under different sensor spectral responses.

This paper proposes a novel iterative maximum weighted independent set (MWIS) algorithm for multiple hypothesis tracking (MHT) in a tracking-by-detection framework. MHT converts the tracking problem into a series of MWIS problems across the tracking time. Previous works solve these NP-hard MWIS problems independently without the use of any prior information from each frame, and they ignore the relevance between adjacent frames. In this paper, we iteratively solve the MWIS problems by using the MWIS solution from the previous frame rather than solving the problem from scratch each time. First, we define five hypothesis categories and a hypothesis transfer model, which explicitly describes the hypothesis relationship between adjacent frames. We also propose a polynomial-time approximation algorithm for the MWIS problem in MHT. In addition to that, we present a confident short tracklet generation method and incorporate tracklet-level association into MHT, which further improves the computational efficiency. Our experiments on both MOT16 and MOT17 benchmarks show that our tracker outperforms all the previously published tracking algorithms on both MOT16 and MOT17 benchmarks. Finally, we demonstrate that the polynomial-time approximate tracker reaches nearly the same tracking performance.

Gaze estimation, which aims to predict gaze points with given eye images, is an important task in computer vision because of its applications in human visual attention understanding. Many existing methods are based on a single camera, and most of them only focus on either the gaze point estimation or gaze direction estimation. In this paper, we propose a novel multitask method for the gaze point estimation using multiview cameras. Specifically, we analyze the close relationship between the gaze point estimation and gaze direction estimation, and we use a partially shared convolutional neural networks architecture to simultaneously estimate the gaze direction and gaze point. Furthermore, we also introduce a new multiview gaze tracking data set that consists of multiview eye images of different subjects. As far as we know, it is the largest multiview gaze tracking data set. Comprehensive experiments on our multiview gaze tracking data set and existing data sets demonstrate that our multiview multitask gaze point estimation solution consistently outperforms existing methods.

We present a novel 3D face reconstruction technique that leverages sparse photometric stereo (PS) and latest advances on face registration / modeling from a single image. We observe that 3D morphable faces approach [21] provides a reasonable geometry proxy for light position calibration. Specifically, we develop a robust optimization technique that can calibrate per-pixel lighting direction and illumination at a very high precision without assuming uniform surface albedos. Next, we apply semantic segmentation on input images and the geometry proxy to refine hairy vs. bare skin regions using tailored filter. Experiments on synthetic and real data show that by using a very small set of images, our technique is able to reconstruct fine geometric details such as wrinkles, eyebrows, whelks, pores, etc, comparable to and sometimes surpassing movie quality productions.

With the rise of misinformation spread via social media channels, enabled by the increasing automation and realism of image manipulation tools, image forensics is an increasingly relevant problem. Classic image forensic methods leverage low-level cues such as metadata, sensor noise fingerprints, and others that are easily fooled when the image is re-encoded upon upload to facebook, etc. This necessitates the use of higher-level physical and semantic cues that, once hard to estimate reliably in the wild, have become more effective due to the increasing power of computer vision. In particular, we detect manipulations introduced by artificial blurring of the image, which creates inconsistent photometric relationships between image intensity and various cues. We achieve 98% accuracy on the most challenging cases in a new dataset of blur manipulations, where the blur is geometrically correct and consistent with the scene's physical arrangement. Such manipulations are now easily generated, for instance, by smartphone cameras having hardware to measure depth, e.g. 'Portrait Mode' of the iPhone7Plus. We also demonstrate good performance on a challenge dataset evaluating a wider range of manipulations in imagery representing 'in the wild' conditions.

The availability of affordable 3D full body reconstruction systems has given rise to free-viewpoint video (FVV) of human shapes. Most existing solutions produce temporally uncorrelated point clouds or meshes with unknown point/vertex correspondences. Individually compressing each frame is ineffective and still yields to ultra-large data sizes. We present an end-to-end deep learning scheme to establish dense shape correspondences and subsequently compress the data. Our approach uses sparse set of "panoramic" depth maps or PDMs, each emulating an inward-viewing concentric mosaics (CM) [45]. We then develop a learning-based technique to learn pixel-wise feature descriptors on PDMs. The results are fed into an autoencoder-based network for compression. Comprehensive experiments demonstrate our solution is robust and effective on both public and our newly captured datasets.

This paper explores gaze prediction in dynamic 360° immersive videos, i.e., based on the history scan path and VR contents, we predict where a viewer will look at an upcoming time. To tackle this problem, we first present the large-scale eye-tracking in dynamic VR scene dataset. Our dataset contains 208 360° videos captured in dynamic scenes, and each video is viewed by at least 31 subjects. Our analysis shows that gaze prediction depends on its history scan path and image contents. In terms of the image contents, those salient objects easily attract viewers' attention. On the one hand, the saliency is related to both appearance and motion of the objects. Considering that the saliency measured at different scales is different, we propose to compute saliency maps at different spatial scales: the sub-image patch centered at current gaze point, the sub-image corresponding to the Field of View (FoV), and the panorama image. Then we feed both the saliency maps and the corresponding images into a Convolutional Neural Network (CNN) for feature extraction. Meanwhile, we also use a Long-Short-Term-Memory (LSTM) to encode the history scan path. Then we combine the CNN features and LSTM features for gaze displacement prediction between gaze point at a current time and gaze point at an upcoming time. Extensive experiments validate the effectiveness of our method for gaze prediction in dynamic VR scenes.

We describe a system that automatically extracts 3D geometry of an indoor scene from a single 2D panorama. Our system recovers the spatial layout by finding the floor, walls, and ceiling; it also recovers shapes of typical indoor objects such as furniture. Using sampled perspective sub-views, we extract geometric cues (lines, vanishing points, orientation map, and surface normals) and semantic cues (saliency and object detection information). These cues are used for ground plane estimation and occlusion reasoning. The global spatial layout is inferred through a constraint graph on line segments and planar superpixels. The recovered layout is then used to guide shape estimation of the remaining objects using their normal information. Experiments on synthetic and real datasets show that our approach is state-of-the-art in both accuracy and efficiency. Our system can handle cluttered scenes with complex geometry that are challenging to existing techniques.

Abstract Headsets for virtual reality such as head‐mounted displays have become ubiquitous and bring immersive experiences to individual users. People who stand outside the virtual world may want to share the same scenes that are shown on the screen of the headset. It is therefore of great importance to merge real and virtual worlds into the same environment, where physical and virtual objects exist simultaneously and interact in real time. We propose shared augmented virtual environment (SAVE), a mixed reality (MR) system that overlays the virtual world with real objects captured by a Kinect depth camera. We refine the depth map and exploit a Graphics Processing Unit (GPU) based natural image matting method to obtain the real objects from cluttered scenes. In the synthetic MR world, we can render real and virtual objects in real time and handle the depth from both worlds properly. The advantage of our system is that we connect the virtual and real worlds with a bridge controller mounted on the Kinect and need to calibrate the whole system only once before use. Our results demonstrate that the proposed SAVE system is able to create high‐quality 1080p live MR footage, enabling realistic virtual experiences to be shared among a number of people in potential applications such as education, design, and entertainment.

In multi-view human body capture systems, the recovered 3D geometry or even the acquired imagery data can be heavily corrupted due to occlusions, noise, limited fieldof- view, etc. Direct estimation of 3D pose, body shape or motion on these low-quality data has been traditionally challenging.In this paper, we present a graph-based non-rigid shape registration framework that can simultaneously recover 3D human body geometry and estimate pose/motion at high fidelity.Our approach first generates a global full-body template by registering all poses in the acquired motion sequence.We then construct a deformable graph by utilizing the rigid components in the global template. We directly warp the global template graph back to each motion frame in order to fill in missing geometry. Specifically,we combine local rigidity and temporal coherence constraints to maintain geometry and motion consistencies. Comprehensive experiments on various scenes show that our method is accurate and robust even in the presence of drastic motions.

Recent advances on image manipulation techniques have made image forgery detection increasingly more challenging. An important component in such tools is to fake motion and/or defocus blurs through boundary splicing and copy-move operators, to emulate wide aperture and slow shutter effects. In this paper, we present a new technique based on the analysis of the camera response functions (CRF) for efficient and robust splicing and copy-move forgery detection and localization. We first analyze how non-linear CRFs affect edges in terms of the intensity-gradient bivariable histograms. We show distinguishable shape differences on real vs. forged blurs near edges after a splicing operation. Based on our analysis, we introduce a deep-learning framework to detect and localize forged edges. In particular, we show the problem can be transformed to a handwriting recognition problem an resolved by using a convolutional neural network. We generate a large dataset of forged images produced by splicing followed by retouching and comprehensive experiments show our proposed method outperforms the state-of-the-art techniques in accuracy and robustness.

We present a novel light field structure-from-motion (SfM) framework for reliable 3D object reconstruction. Specifically, we use the light field (LF) camera such as Lytro and Raytrix as a virtual 3D scanner. We move an LF camera around the object and register between multiple LF shots. We show that applying conventional SfM on sub-aperture images is not only expensive but also unreliable due to ultra-small baseline and low image resolution. Instead, our LF-SfM scheme maps ray manifolds across LFs. Specifically, we show how rays passing through a common 3D point transform between two LFs and we develop reliable technique for extracting extrinsic parameters from this ray transform. Next, we apply a new edge-preserving stereo matching technique on individual LFs and conduct LF bundle adjustment to jointly optimize pose and geometry. Comprehensive experiments show our solution outperforms many state-of-the-art passive and even active techniques especially on topologically complex objects.

The complete plenoptic function records radiance of rays from every location, at every angle, for every wavelength and at every time. The signal is multi-dimensional and has long relied on multi-modal sensing such as hybrid light field camera arrays. In this paper, we present a single camera hyperspectral light field imaging solution that we call Snapshot Plenoptic Imager (SPI). SPI uses spectral coded catadioptric mirror arrays for simultaneously acquiring the spatial, angular and spectral dimensions. We further apply a learning-based approach to improve the spectral resolution from very few measurements. Specifically, we demonstrate and then employ a new spectral sparsity prior that allows the hyperspectral profiles to be sparsely represented under a pre-trained dictionary. Comprehensive experiments on synthetic and real data show that our technique is effective, reliable, and accurate. In particular, we are able to produce the first wide FoV multi-spectral light field database.

Traditional Structure-from-Motion (SfM) uses images captured by cameras as inputs. In this paper, we explore using light fields captured by plenoptic cameras or camera arrays as inputs. We call this solution plenoptic SfM or P-SfM solution. We first present a comprehensive theory on ray geometry transforms under light field pose variations. We derive the transforms of three typical ray manifolds: rays passing through a point or point-ray manifold, rays passing through a 3D line or ray-line manifold, and rays lying on a common 3D plane or ray-plane manifold. We show that by matching these manifolds across LFs, we can recover light field poses and conduct bundle adjustment in ray space. We validate our theory and framework on synthetic and real data on light fields of different scales: small scale LFs acquired using a LF camera and large scale LFs by a camera array. We show that our P-SfM technique can significantly improve the accuracy and reliability over regular SfM and PnP especially on traditionally challenging scenes where reliable feature point correspondences are difficult to obtain but line or plane correspondences are readily accessible.

Saliency detection is a long standing problem in computer vision. Tremendous efforts have been focused on exploring a universal saliency model across users despite their differences in gender, race, age, etc. Yet recent psychology studies suggest that saliency is highly specific than universal: individuals exhibit heterogeneous gaze patterns when viewing an identical scene containing multiple salient objects. In this paper, we first show that such heterogeneity is common and critical for reliable saliency prediction. Our study also produces the first database of personalized saliency maps (PSMs). We model PSM based on universal saliency map (USM) shared by different participants and adopt a multi-task CNN framework to estimate the discrepancy between PSM and USM. Comprehensive experiments demonstrate that our new PSM model and prediction scheme are effective and reliable.

State-of-the-art calibration methods typically choose to use a checkerboard as the calibration target for its simplicity and robustness. They however require the complete checkerboard be captured to break symmetry. More recent multi-camera systems such as Google Jump, Jaunt, and camera arrays have limited overlapping field-of-view (FoV) and having all cameras viewing the complete checkerboard is extremely difficult in reality. Tailored patterns such as CALTag [1] introduce new image features within the checker blocks for breaking symmetry but they also break the grid topology. We present a new technique using such patterned calibration targets for a broad range of multi-camera systems. Our key observation is that applying directional gradient filters yields to heterogeneous responses on grid vs. non-grid features: the former are isolated and the latter are highly inter-connected. We therefore apply a simple but highly efficient technique to eliminate non-grid outliers based on connected component analysis and gradient histograms. Finally, we recover the complete grid by approximating each local checkerboard as a parallelogram and imposing the topology constraint. We conduct comprehensive experiments on a number of recent multi-camera systems and our technique significantly outperforms the state-of-the-art in accuracy and robustness.

Light field imaging is an emerging technology in computational photography areas. Based on innovative designs of the imaging model and the optical path, light field cameras not only record the spatial intensity of threedimensional (3D) objects, but also capture the angular information of the physical world, which provides new ways to address various problems in computer vision, such as 3D reconstruction, saliency detection, and object recognition. In this paper, three key aspects of light field cameras, i.e., model, calibration, and reconstruction, are reviewed extensively. Furthermore, light field based applications on informatics, physics, medicine, and biology are exhibited. Finally, open issues in light field imaging and long-term application prospects in other natural sciences are discussed.

Existing saliency detection approaches use images as inputs and are sensitive to foreground/background similarities, complex background textures, and occlusions. We explore the problem of using light fields as input for saliency detection. Our technique is enabled by the availability of commercial plenoptic cameras that capture the light field of a scene in a single shot. We show that the unique refocusing capability of light fields provides useful focusness, depths, and objectness cues. We further develop a new saliency detection algorithm tailored for light fields. To validate our approach, we acquire a light field database of a range of indoor and outdoor scenes and generate the ground truth saliency map. Experiments show that our saliency detection scheme can robustly handle challenging scenarios such as similar foreground and background, cluttered background, complex occlusions, etc., and achieve high accuracy and robustness.

When using plenoptic camera for digital refocusing, angular undersampling can cause severe (angular) aliasing artifacts. Previous approaches have focused on avoiding aliasing by pre-processing the acquired light field via prefiltering, demosaicing, reparameterization, and so on. In this paper, we present a different solution that first detects and then removes angular aliasing at the light field refocusing stage. Different from previous frequency domain aliasing analysis, we carry out a spatial domain analysis to reveal whether the angular aliasing would occur and uncover where in the image it would occur. The spatial analysis also facilitates easy separation of the aliasing versus non-aliasing regions and angular aliasing removal. Experiments on both synthetic scene and real light field data sets (camera array and Lytro camera) demonstrate that our approach has a number of advantages over the classical prefiltering and depth-dependent light field rendering techniques.

The availability of commodity multi-camera systems such as Google Jump, Jaunt, and Lytro Immerge have brought new demand for reliable and efficient extrinsic camera calibration. State-of-the-art solutions generally require that adjacent, if not all, cameras observe a common area or employ known scene structures. In this paper, we present a novel multi-camera calibration technique that eliminates such requirements. Our approach extends the single-pair hand-eye calibration used in robotics to multi-camera systems. Specifically, we make use of (possibly unknown) planar structures in the scene and combine plane-based structure from motion, camera pose estimation, and task-specific bundle adjustment for extrinsic calibration. Experiments on several multi-camera setups demonstrate that our scheme is highly accurate, robust, and efficient.

Light fields (LFs) are image-based representation that records the radiance along all rays along every direction through every point in space. Traditionally LFs are acquired by using a 2D grid of evenly spaced pinhole cameras or by translating a pinhole camera along the 2D grid using a robot arm. In this paper, we present a novel LF sampling scheme by exploiting a special non-centric camera called the crossed-slit or XSlit camera. An XSlit camera acquires rays that simultaneously pass through two oblique slits. We show that, instead of translating the camera as in the pinhole case, we can effectively sample the LF by rotating individual or both slits while keeping the camera fixed. This leads a "fixed-location" LF acquisition scheme. We further show through theoretical analysis and experiments that the resulting XSlit LFs provide several advantages: they provide more dense spatial-angular sampling, are amenable multi-view stereo matching and volumetric reconstruction, and can synthesize unique refocusing effects.

Light fields (LFs) have been shown to enable photorealistic visualization of complex scenes. In practice, however, an LF tends to have a relatively small angular range or spatial resolution, which limits the scope of virtual navigation. In this paper, we show how seamless virtual navigation can be enhanced by stitching multiple LFs. Our technique consists of two key components: LF registration and LF stitching. To register LFs, we use what we call the ray-space motion matrix (RSMM) to establish pairwise ray-ray correspondences. Using Plücker coordinates, we show that the RSMM is a 5 ×6 matrix, which reduces to a 5 ×5 matrix under pure translation and/or in-plane rotation. The final LF stitching is done using multi-resolution, high-dimensional graph-cut in order to account for possible scene motion, imperfect RSMM estimation, and/or undersampling. We show how our technique allows us to create LFs with various enhanced features: extended horizontal and/or vertical field-of-view, larger synthetic aperture and defocus blur, and larger parallax.

There is an emerging interest on using high-dimensional datasets beyond 2D images in saliency detection. Examples include 3D data based on stereo matching and Kinect sensors and more recently 4D light field data. However, these techniques adopt very different solution frameworks, in both type of features and procedures on using them. In this paper, we present a unified saliency detection framework for handling heterogenous types of input data. Our approach builds dictionaries using data-specific features. Specifically, we first select a group of potential foreground superpixels to build a primitive saliency dictionary. We then prune the outliers in the dictionary and test on the remaining superpixels to iteratively refine the dictionary. Comprehensive experiments show that our approach universally outperforms the state-of-the-art solution on all 2D, 3D and 4D data.

There has been emerging interest on recovering traditionally challenging intrinsic scene properties. In this paper, we present a novel computational imaging solution for recovering the ambient occlusion (AO) map of an object. AO measures how much light from all different directions can reach a surface point without being blocked by self-occlusions. Previous approaches either require obtaining highly accurate surface geometry or acquiring a large number of images. We adopt a compressive sensing framework that captures the object under strategically coded lighting directions. We show that this incident illumination field exhibits some unique properties suitable for AO recovery: every ray's contribution to the visibility function is binary while their distribution for AO measurement is sparse. This enables a sparsity-prior based solution for iteratively recovering the surface normal, the surface albedo, and the visibility function from a small number of images. To physically implement the scheme, we construct an encodable directional light source using the light field probe. Experiments on synthetic and real scenes show that our approach is both reliable and accurate with significantly reduced size of input.

We describe a technique to recover depth from a light field (LF) using two proposed features of the LF focal stack. One feature is the property that non-occluding pixels exhibit symmetry along the focal depth dimension centered at the in-focus slice. The other is a data consistency measure based on analysis-by-synthesis, i.e., the difference between the synthesized focal stack given the hypothesized depth map and that from the LF. These terms are used in an iterative optimization framework to extract scene depth. Experimental results on real Lytro and Raytrix data demonstrate that our technique outperforms state-of-the-art solutions and is significantly more robust to noise and under-sampling.

In perspective cameras, images of a frontal-parallel 3D object preserve its aspect ratio invariant to its depth. Such an invariance is useful in photography but is unique to perspective projection. In this paper, we show that alternative non-perspective cameras such as the crossed-slit or XSlit cameras exhibit a different depth-dependent aspect ratio (DDAR) property that can be used to 3D recovery. We first conduct a comprehensive analysis to characterize DDAR, infer object depth from its AR, and model recoverable depth range, sensitivity, and error. We show that repeated shape patterns in real Manhattan World scenes can be used for 3D reconstruction using a single XSlit image. We also extend our analysis to model slopes of lines. Specifically, parallel 3D lines exhibit depth-dependent slopes (DDS) on their images which can also be used to infer their depths. We validate our analyses using real XSlit cameras, XSlit panoramas, and catadioptric mirrors. Experiments show that DDAR and DDS provide important depth cues and enable effective single-image scene reconstruction.

Patch-based denoising, while effective, requires expensive pairwise patch comparisons. We present a novel fast patch-based denoising technique based on patch geodesic paths (PatchGPs). PatchGPs treat image patches as nodes and patch differences as edge weights for computing the shortest (geodesic) paths. The distance defined by the PatchGP can then be used as a similarity metric for image denoising. We first show that, for natural images, PatchGPs can be approximated by minimum hop paths (MHPs) that correspond to Euclidean line paths connecting two patch nodes. The denoising kernel is constructed using patches along discretized MHP search directions. We apply a weight propagation scheme to robustly and efficiently compute the path distance for each MHP. Our technique handles noise at multiple scales by analyzing the noise distribution (through wavelet decomposition) at each scale. Experiments show that our approach maintains the high quality of patch-based denoising but is a few orders of magnitude faster. We also demonstrate how PatchGP can be used for fast Bayer pattern (raw) denoising and image detail enhancement.

In this paper, we introduce a bilateral consistency metric on the surface camera (SCam) [26] for light field stereo matching to handle significant occlusions. The concept of SCam is used to model angular radiance distribution with respect to a 3D point. Our bilateral consistency metric is used to indicate the probability of occlusions by analyzing the SCams. We further show how to distinguish between on-surface and free space, textured and non-textured, and Lambertian and specular through bilateral SCam analysis. To speed up the matching process, we apply the edge preserving guided filter [14] on the consistency-disparity curves. Experimental results show that our technique outperforms both the state-of-the-art and the recent light field stereo matching methods, especially near occlusion boundaries.

Robust tracking of deformable object like catheter or vascular structures in X-ray images is an important technique used in image guided medical interventions for effective motion compensation and dynamic multi-modality image fusion. Tracking of such anatomical structures and devices is very challenging due to large degrees of appearance changes, low visibility of X-ray images and the deformable nature of the underlying motion field as a result of complex 3D anatomical movements projected into 2D images. To address these issues, we propose a new deformable tracking method using the tensor-based algorithm with model propagation. Specifically, the deformable tracking is formulated as a multi-dimensional assignment problem which is solved by rank-1 l1tensor approximation. The model prior is propagated in the course of deformable tracking. Both the higher order information and the model prior provide powerful discriminative cues for reducing ambiguity arising from the complex background, and consequently improve the tracking robustness. To validate the proposed approach, we applied it to catheter and vascular structures tracking and tested on X-ray fluoroscopic sequences obtained from 17 clinical cases. The results show, both quantitatively and qualitatively, that our approach achieves a mean tracking error of 1.4 pixels for vascular structure and 1.3 pixels for catheter tracking.

The goal of image pre-compensation is to process an image such that after being convolved with a known kernel, will appear close to the sharp reference image. In a practical setting, the pre-compensated image has significantly higher dynamic range than the latent image. As a result, some form of tone mapping is needed. In this paper, we show how global tone mapping functions affect contrast and ringing in image pre-compensation. In particular, we show that linear tone mapping eliminates ringing but incurs severe contrast loss, while non-linear tone mapping functions such as Gamma curves slightly enhances contrast but introduces ringing. To enable quantitative analysis, we design new metrics to measure the contrast of an image with ringing. Specifically, we set out to find its "equivalent ringing-free" image that matches its intensity histogram and uses its contrast as the measure. We illustrate our approach on projector defocus compensation and visual acuity enhancement. Compared with the state-of-the-art, our approach significantly improves the contrast. We believe our technique is the first to analytically trade-off between contrast and ringing.

Existing saliency detection approaches use images as inputs and are sensitive to foreground/background similarities, complex background textures, and occlusions. We explore the problem of using light fields as input for saliency detection. Our technique is enabled by the availability of commercial plenoptic cameras that capture the light field of a scene in a single shot. We show that the unique refocusing capability of light fields provides useful focusness, depths, and objectness cues. We further develop a new saliency detection algorithm tailored for light fields. To validate our approach, we acquire a light field database of a range of indoor and outdoor scenes and generate the ground truth saliency map. Experiments show that our saliency detection scheme can robustly handle challenging scenarios such as similar foreground and background, cluttered background, complex occlusions, etc, and achieve high accuracy and robustness.

When using plenoptic camera for digital refocusing, angular undersampling can cause severe (angular) aliasing artifacts. Previous approaches have focused on avoiding aliasing by pre-processing the acquired light field via prefiltering, demosaicing, reparameterization, etc. In this paper, we present a different solution that first detects and then removes aliasing at the light field refocusing stage. Different from previous frequency domain aliasing analysis, we carry out a spatial domain analysis to reveal whether the aliasing would occur and uncover where in the image it would occur. The spatial analysis also facilitates easy separation of the aliasing vs. non-aliasing regions and aliasing removal. Experiments on both synthetic scene and real light field camera array data sets demonstrate that our approach has a number of advantages over the classical prefiltering and depth-dependent light field rendering techniques.

Generating realistic hairstyles is critical in modeling human characters in computer games and feature films. State-of-the-art geometric modeling tools demand extensive user efforts and artistic skills. Image-based approaches attempt to automate the modeling process but require using special acquisition systems and/or computer vision techniques. In this paper, we present a new hybrid image-CAD based hairstyle modeling tool. We use the rich resource of hairstyle photos from fashion magazines as inputs where each hairstyle is commonly captured with a high quality and from multiple (2 ~ 3) viewpoints. We develop a sketch-based interface to allow users to draw a small number of strokes on different views and then map them to a sparse set of 3D splines. We then connect a web of splines as a guide hair surface and use it to compute a 3D orientational field within the hair volume for growing hair fibers. To further refine hairstyle models, we provide a set of tools for adjusting hair fiber length and density, perturbing hair fibers, creating wisps and curls, etc. Finally, we integrate an interactive rendering engine with the modeling modules so that the user can assess the model under realistic shading. Compared with existing solutions, our system significantly reduces user efforts and produces high quality hairstyle models consistent with the underlying hairstyle in the photo.

Patch-based methods such as Non-Local Means (NLM) and BM3D have become the de facto gold standard for image denoising. The core of these approaches is to use similar patches within the image as cues for denoising. The operation usually requires expensive pair-wise patch comparisons. In this paper, we present a novel fast patch-based denoising technique based on Patch Geodesic Paths (PatchGP). PatchGPs treat image patches as nodes and patch differences as edge weights for computing the shortest (geodesic) paths. The path lengths can then be used as weights of the smoothing/denoising kernel. We first show that, for natural images, PatchGPs can be effectively approximated by minimum hop paths (MHPs) that generally correspond to Euclidean line paths connecting two patch nodes. To construct the denoising kernel, we further discretize the MHP search directions and use only patches along the search directions. Along each MHP, we apply a weight propagation scheme to robustly and efficiently compute the path distance. To handle noise at multiple scales, we conduct wavelet image decomposition and apply PatchGP scheme at each scale. Comprehensive experiments show that our approach achieves comparable quality as the state-of-the-art methods such as NLM and BM3D but is a few orders of magnitude faster.

Transparent gas flows are difficult to reconstruct: the refractive index field (RIF) within the gas volume is uneven and rapidly evolving, and correspondence matching under distortions is challenging. We present a novel computational imaging solution by exploiting the light field probe (LF-Probe). A LF-probe resembles a view-dependent pattern where each pixel on the pattern maps to a unique ray. By observing the LF-probe through the gas flow, we acquire a dense set of ray-ray correspondences and then reconstruct their light paths. To recover the RIF, we use Fermat's Principle to correlate each light path with the RIF via a Partial Differential Equation (PDE). We then develop an iterative optimization scheme to solve for all light-path PDEs in conjunction. Specifically, we initialize the light paths by fitting Hermite splines to ray-ray correspondences, discretize their PDEs onto voxels, and solve a large, over-determined PDE system for the RIF. The RIF can then be used to refine the light paths. Finally, we alternate the RIF and light-path estimations to improve the reconstruction. Experiments on synthetic and real data show that our approach can reliably reconstruct small to medium scale gas flows. In particular, when the flow is acquired by a small number of cameras, the use of ray-ray correspondences can greatly improve the reconstruction.

A Manhattan World (MW) [3] is composed of planar surfaces and parallel lines aligned with three mutually orthogonal principal axes. Traditional MW understanding algorithms rely on geometry priors such as the vanishing points and reference (ground) planes for grouping coplanar structures. In this paper, we present a novel single-image MW reconstruction algorithm from the perspective of non-pinhole cameras. We show that by acquiring the MW using an XSlit camera, we can instantly resolve co planarity ambiguities. Specifically, we prove that parallel 3D lines map to 2D curves in an XSlit image and they converge at an XSlit Vanishing Point (XVP). In addition, if the lines are coplanar, their curved images will intersect at a second common pixel that we call Coplanar Common Point (CCP). CCP is a unique image feature in XSlit cameras that does not exist in pinholes. We present a comprehensive theory to analyze XVPs and CCPs in a MW scene and study how to recover 3D geometry in a complex MW scene from XVPs and CCPs. Finally, we build a prototype XSlit camera by using two layers of cylindrical lenses. Experimental results on both synthetic and real data show that our new XSlit-camera-based solution provides an effective and reliable solution for MW understanding.

The goal of saliency detection is to locate important pixels or regions in an image which attract humans' visual attention the most. This is a fundamental task whose output may serve as the basis for further computer vision tasks like segmentation, resizing, tracking and so forth. In this paper we propose a novel salient region detection algorithm by integrating three important visual cues namely uniqueness, focus ness and objectness (UFO). In particular, uniqueness captures the appearance-derived visual contrast, focus ness reflects the fact that salient regions are often photographed in focus, and objectness helps keep completeness of detected salient regions. While uniqueness has been used for saliency detection for long, it is new to integrate focus ness and objectness for this purpose. In fact, focus ness and objectness both provide important saliency information complementary of uniqueness. In our experiments using public benchmark datasets, we show that, even with a simple pixel level combination of the three components, the proposed approach yields significant improvement compared with previously reported methods.

Traditional stereo matching assumes perspective viewing cameras under a translational motion: the second camera is translated away from the first one to create parallax. In this paper, we investigate a different, rotational stereo model on a special multi-perspective camera, the XSlit camera. We show that rotational XSlit (R-XSlit) stereo can be effectively created by fixing the sensor and slit locations but switching the two slits' directions. We first derive the epipolar geometry of R-XSlit in the 4D light field ray space. Our derivation leads to a simple but effective scheme for locating corresponding epipolar "curves". To conduct stereo matching, we further derive a new disparity term in our model and develop a patch-based graph-cut solution. To validate our theory, we assemble an XSlit lens by using a pair of cylindrical lenses coupled with slit-shaped apertures. The XSlit lens can be mounted on commodity cameras where the slit directions are adjustable to form desirable R-XSlit pairs. We show through experiments that R-XSlit provides a potentially advantageous imaging system for conducting fixed-location, dynamic baseline stereo.

Light fields are image-based representations that use densely sampled rays as a scene description. In this paper, we explore geometric structures of 3D lines in ray space for improving light field triangulation and stereo matching. The triangulation problem aims to fill in the ray space with continuous and non-overlapping simplices anchored at sampled points (rays). Such a triangulation provides a piecewise-linear interpolant useful for light field super-resolution. We show that the light field space is largely bilinear due to 3D line segments in the scene, and direct triangulation of these bilinear subspaces leads to large errors. We instead present a simple but effective algorithm to first map bilinear subspaces to line constraints and then apply Constrained Delaunay Triangulation (CDT). Based on our analysis, we further develop a novel line-assisted graph-cut (LAGC) algorithm that effectively encodes 3D line constraints into light field stereo matching. Experiments on synthetic and real data show that both our triangulation and LAGC algorithms outperform state-of-the-art solutions in accuracy and visual quality.

This paper investigates the role that nonlinear camera response functions (CRFs) have on image deblurring. We present a comprehensive study to analyze the effects of CRFs on motion deblurring. In particular, we show how nonlinear CRFs can cause a spatially invariant blur to behave as a spatially varying blur. We prove that such nonlinearity can cause large errors around edges when directly applying deconvolution to a motion blurred image without CRF correction. These errors are inevitable even with a known point spread function (PSF) and with state-of-the-art regularization-based deconvolution algorithms. In addition, we show how CRFs can adversely affect PSF estimation algorithms in the case of blind deconvolution. To help counter these effects, we introduce two methods to estimate the CRF directly from one or more blurred images when the PSF is known or unknown. Our experimental results on synthetic and real images validate our analysis and demonstrate the robustness and accuracy of our approaches.

This paper presents a novel coprime blurred pair (CBP) model to improve data security in camera surveillance. While most previous approaches have focused on completely encrypting the video stream, we introduce a spatial encryption scheme by strategically blurring the image/video contents. Specifically, we form a public stream and a private stream by blurring the original video data using two different kernels. Each blurred stream will provide the user who has lower clearance less access to personally identifiable details while still allowing behavior to be monitored. If the behavior is recognized as suspicious, a supervisor can use both streams to deblur the contents. Our approach is based on a new CBP theory where the two kernels are coprime when mapped to bivariate polynomials in the $(z)$ domain. We show that coprimality can be derived in terms of the rank of Bézout matrix formed by sampled polynomials, and we present an efficient algorithm to factor the Bézout matrix for recovering the latent image. To make our solution practical, we implement our decryption scheme on a graphics processing unit (GPU) to achieve real-time performance. Extensive experiments demonstrate that our new scheme can effectively protect sensitive identity information in surveillance videos and faithfully reconstruct the unblurred video stream when both CBP sequences are available.

Figures play an important role in illustrating concepts, methodology and results in biomedicai literature. However, figures in biomedicai literature are often composed of multiple subfigures (panels), which may illustrate diverse methodologies or results. Robust and accurate panel partitioning is crucial to support article categorization based on methods or experimental results and to provide the evidence source for derived assertions. But, it is a challenging task. In this paper, we present a comprehensive framework for harvesting multimodal panels in biomedicai literature, and demonstrate its application to protein-protein interaction (PPI)-related literature as a use case. A unique feature of our solution is that we combine pixel-level representations of images with figure captions. Our approach first analyzes figure captions to identify the label style used to mark panels. We then use pixel-level representations to partition a figure into a set of bounding boxes of connected components. We also perform a lexical analysis on the text within the figure to locate panel labels that match the caption analysis results. Finally, we estimate the optimal panel layout and use the layout to partition the figure. We tested our system on a dataset provided by the Molecular INTeraction database (MINT), and show that our approach surpasses pure caption-based and pure image-based approaches, achieving a 96.64% precision.

We present a novel and simple computational imaging solution to robustly and accurately recover 3D dynamic fluid surfaces. Traditional specular surface reconstruction schemes place special patterns (checkerboard or color patterns) beneath the fluid surface to establish point-pixel correspondences. However, point-pixel correspondences alone are insufficient to recover surface normal or height and they rely on additional constraints to resolve the ambiguity. In this paper, we exploit using Bokode - a computational optical device that emulates a pinhole projector - for capturing ray-ray correspondences which can then be used to directly recover the surface normals. We further develop a robust feature matching algorithm based on the Active-Appearance Model to robustly establishing ray-ray correspondences. Our solution results in an angularly sampled normal field and we derive a new angular-domain surface integration scheme to recover the surface from the normal fields. Specifically, we reformulate the problem as an over-constrained linear system under spherical coordinate and solve it using Singular Value Decomposition. Experiments results on real and synthetic surfaces demonstrate that our approach is robust and accurate, and is easier to implement than state-of-the-art multi-camera based approaches.

A plenoptic camera captures the 4D radiance about a scene. Recent practical solutions mount a microlens array on top of a commodity SLR to directly acquire these rays. However, they suffer from low resolution as hundreds of thousands of views need to be captured in a single shot. In this paper, we develop a simple but effective technique for improving the image resolution of the plenoptic camera by maneuvering the demosaicing process. We first show that the traditional solution by demosaicing each individual microlens image and then blending them for view synthesis is suboptimal. In particular, this demosaicing process often suffers from aliasing artifacts, and it damages high frequency information recorded by each microlens image hence degrades the image quality. We instead propose to de-mosaic the synthesized view at the rendering stage. Specifically, we first transform the radiance to the desired focal plane and then apply frequency domain plenoptic resampling. A full resolution color filtered image is then created by performing a 2D integral projection from the reparam-eterized radiance. Finally, we conduct demosacing to obtain the color result. We show that our solution can achieve visible resolution enhancement on dynamic refocusing and depth-assisted deep focus rendering.

The appearance of hair plays a critical role in synthesizing realistic looking human characters. However, due to the high complexity in hair geometry and the scattering nature of hair fibers, rendering hair with photorealistic quality and at interactive speeds remains as an open problem in computer graphics. Previous approaches attempt to simplify the scattering model to only tackle a specific aspect of the scattering effects. In this paper, we present a new approach to simultaneously render complex scattering effects including volumetric shadows, transparency, and antialiasing under a unified framework. Our solution uses a shadow-ray path to produce volumetric self-shadows and an additional view-ray path to produce transparency. To compute and accumulate the contribution of individual hair fibers along each (shadow or view) path, we develop a new GPU-based k-buffer technique that can efficiently locate the K nearest scattering locations and combine them in the correct order. Compared with existing multi-layer based approaches[Kim and Neumann 2001; Yuksel and Keyser 2008; Sintorn and Assarsson 2009], we show that our k-buffer solution can more accurately reproduce the shadowing and transparency effects. Further, we present an anti-aliasing scheme that directly builds upon the k-buffer. We implement all three effects (volumetric shadows, transparency, and anti-aliasing) under a unified rendering pipeline. Experiments on complex hair models demonstrate that our new solution produces near photorealistic hair rendering at very interactive speed.

We address the problem of motion deblurring using coded exposure. This approach allows for accurate estimation of a sharp latent image via well-posed deconvolution and avoids lost image content that cannot be recovered from images acquired with a traditional shutter. Previous work in this area has used either manual user input or alpha matting approaches to estimate the coded exposure Point Spread Function (PSF) from the captured image. In order to automate deblurring and to avoid the limitations of matting approaches, we propose a Fourier-domain statistical approach to coded exposure PSF estimation that allows us to estimate the latent image in cases of constant velocity, constant acceleration, and harmonic motion. We further demonstrate that previously used criteria to choose a coded exposure PSF do not produce one with optimal reconstruction error, and that an additional 30 percent reduction in Root Mean Squared Error (RMSE) of the latent image estimate can be achieved by incorporating natural image statistics.

Figures in biomedical articles often constitute direct evidence of experimental results. Image analysis methods can be coupled with text-based methods to improve knowledge discovery. However, automatically harvesting figures along with their associated captions from full-text articles remains challenging. In this paper, we present an automatic system for robustly harvesting figures from biomedical literature. Our approach relies on the idea that the PDF specification of the document layout can be used to identify encoded figures and figure boundaries within the PDF and enforce constraints among figure-regions. This allows us to harvest fragments of figures (subflgures), from the PDF, correctly identify subfigures that belong to the same figure, and identify the captions associated with each figure. Our method simultaneously recovers figures and captions and applies additional filtering process to remove irrelevant figures such as logos, to eliminate text passages that were incorrectly identified as captions, and to re-group subflgures to generate a putative figure. Finally, we associate figures with captions. Our preliminary experiments suggest that our method achieves an accuracy of 95% in harvesting figures-caption pairs from a set of 2,035 full-text biomedical documents from BioCreative III, containing 12,574 figures.

Content-aware image retargeting has attracted a lot of interests recently. The key and most challenging issue for this task is how to balance the tradeoff between preserving the important contents and minimizing the visual distortions on the consistency of the image structure. In this paper we present a novel filtering-based technique to tackle this issue, called ”importance filtering”. Specifically, we first filter the image saliency guided by the image itself to achieve a structure-consistent importance map. We then use the pixel importance as the key constraint to compute the gradient map of pixel shifts from the original resolution to the target. Finally, we integrate the shift gradient across the image using a weighted filter to construct a smooth shift map and render the target image. The weight is again controlled by the pixel importance. The two filtering processes enforce to maintain the structural consistency and yet preserve the important contents in the target image. Furthermore, the simple nature of filter operations allows highly efficient implementation for real-time applications and easy extension to video retargeting, as the structural constraints from the original image naturally convey the temporal coherence between frames. The effectiveness and efficiency of our importance filtering algorithm are confirmed in extensive experiments.

We present a novel theory for characterizing defocus blurs in multi-perspective cameras such as catadioptric mirrors. Our approach studies how multi-perspective ray geometry transforms under the thin lens. We first use the General Linear Cameras (GLCs) to approximate the incident multi-perspective rays to the lens and then apply a Thin Lens Operator (TLO) to map an incident GLC to the exit GLC. To study defocus blurs caused by the GLC rays, we further introduce a new Ray Spread Function (RSF) model analogous the Point Spread Function (PSF). While PSF models defocus blurs caused by a 3D scene point, RSF models blurs spread by rays. We derive closed form RSFs for incident GLC rays, and we show that for catadioptric cameras with a circular aperture, the RSF can be effectively approximated as a single or mixtures of elliptic-shaped kernels. We apply our method for predicting defocus blurs on commonly used catadioptric cameras and for reducing de-focus blurs in catadioptric projections. Experiments on synthetic and real data demonstrate the accuracy and general applicability of our approach.

Acquiring dynamic 3D fluid surfaces is a challenging problem in computer vision. Single or stereo camera based solutions are sensitive to refraction distortions, fast fluid motions, and calibration errors. In this paper, we present a multi-view based solution for robustly capturing fast evolving fluid wavefronts. We first construct a portable, 3×3 camera array system as the main acquisition device. We elaborately design the system to allow high-resolution and high-speed capture. To recover fluid surfaces, we place a known pattern beneath the surface and position the camera array on top to observe the pattern. By tracking the distorted feature points over time and across cameras, we obtain spatial-temporal correspondence maps and we use them for specular carving to reconstruct the time-varying surface. In case one of the cameras loses track due to distortions or blurs, we use the rest of the cameras to construct the surface and then apply multi-perspective warping to locate the lost-track feature points so that we can continue using the camera in later frames. Our experiments on synthetic and real data demonstrate that our multi-view framework is robust and reliable.

We present a new Coprime Blurred Pair (CBP) theory that may benefit a number of computer vision applications. A CBP is constructed by blurring the same latent image with two unknown kernels, where the two kernels are co-prime when mapped to bivariate polynomials under the z-transform. We first show that the blurred contents in a CBP are difficult to restore using conventional blind deconvolution methods based on sparsity priors. We therefore introduce a new coprime prior for recovering the latent image in a CBP. Our solution maps the CBP to bivariate polynomials and sample them on the unit circle in both dimension. We show that coprimality can be derived in terms of the rank of the Bézout Matrix [2] formed by the sampled polynomials and we present an efficient algorithm to factor the Bézout Matrix for recovering the latent image. Finally, we discuss applications of the CBP theory in privacy-preserving surveillance and motion deblurring, as well as physical implementations of CBPs using flutter shutter cameras.

Visual tracking plays an important role in many computer vision tasks. A common assumption in previous methods is that the video frames are blur free. In reality, motion blurs are pervasive in the real videos. In this paper we present a novel BLUr-driven Tracker (BLUT) framework for tracking motion-blurred targets. BLUT actively uses the information from blurs without performing debluring. Specifically, we integrate the tracking problem with the motion-from-blur problem under a unified sparse approximation framework. We further use the motion information inferred by blurs to guide the sampling process in the particle filter based tracking. To evaluate our method, we have collected a large number of video sequences with significant motion blurs and compared BLUT with state-of-the-art trackers. Experimental results show that, while many previous methods are sensitive to motion blurs, BLUT can robustly and reliably track severely blurred targets.

The classical optical flow assumes that a feature point maintains constant brightness across the frames. For fluid-type motions such as smoke or clouds, the constant brightness assumption does not hold, and accurately estimating the motion flow from their images is difficult. In this paper, we introduce a simple but effective Navier-Stokes (NS) potential flow model for recovering fluid-type motions. Our method treats the image as a wavefront surface and models the 3D potential flow beneath the surface. The gradient of the velocity potential describes the motion flow at every voxel. We first derive a general brightness constraint that explicitly models wavefront (brightness) variations in terms of the velocity potential. We then use a series of partial differential equations to separately model the dynamics of the potential flow. To solve for the potential flow, we use the Dirichlet-Neumann Operator (DNO) to simplify the 3D volumetric velocity potential to 2D surface velocity potential. We approximate the DNO via Taylor expansions and develop a Fourier domain method to efficiently estimate the Taylor coefficients. Finally we show how to use the DNO to recover the velocity potential from images as well as to propagate the wavefront (image) over time. Experimental results on both synthetic and real images show that our technique is robust and reliable.

Abstract We present a novel image‐based technique for modeling complex unfoliaged trees. Existing tree modeling tools either require capturing a large number of views for dense 3D reconstruction or rely on user inputs and botanic rules to synthesize natural‐looking tree geometry. In this paper, we focus on faithfully recovering real instead of realistically‐looking tree geometry from a sparse set of images. Our solution directly integrates 2D/3D tree topology as shape priors into the modeling process. For each input view, we first estimate a 2D skeleton graph from its matte image and then find a 2D skeleton tree from the graph by imposing tree topology. We develop a simple but effective technique for computing the optimal 3D skeleton tree most consistent with the 2D skeletons. For each edge in the 3D skeleton tree, we further apply volumetric reconstruction to recover its corresponding curved branch. Finally, we use piecewise cylinders to approximate each branch from the volumetric results. We demonstrate our framework on a variety of trees to illustrate the robustness and usefulness of our technique.

Abstract A perspective image represents the spatial relationships of objects in a scene as they appear from a single viewpoint. In contrast, a multi‐perspective image combines what is seen from several viewpoints into a single image. Despite their incongruity of view, effective multi‐perspective images are able to preserve spatial coherence and can depict, within a single context, details of a scene that are simultaneously inaccessible from a single view, yet easily interpretable by a viewer. In computer vision, multi‐perspective images have been used for analysing structure revealed via motion and generating panoramic images with a wide field‐of‐view using mirrors. In this STAR, we provide a practical guide on topics in multi‐perspective modelling and rendering methods and multi‐perspective imaging systems. We start with a brief review of multi‐perspective image techniques frequently employed by artists such as the visual paradoxes of Escher, the Cubism of Picasso and Braque and multi‐perspective panoramas in cel animations. We then characterize existing multi‐perspective camera models, with an emphasis on their underlying geometry and image properties. We demonstrate how to use these camera models for creating specific multi‐perspective rendering effects. Furthermore, we show that many of these cameras satisfy the multi‐perspective stereo constraints and we demonstrate several multi‐perspective imaging systems for extracting 3D geometry for computer vision. The participants learn about topics in multi‐perspective modelling and rendering for generating compelling pictures for computer graphics and in multi‐perspective imaging for extracting 3D geometry for computer vision. We hope to provide enough fundamentals to satisfy the technical specialist without intimidating curious digital artists interested in multi‐perspective images. The intended audience includes digital artists, photographers and computer graphics and computer vision researchers using or building multi‐perspective cameras. They will learn about multi‐perspective modelling and rendering, along with many real world multi‐perspective imaging systems.

Abstract We present a new algorithm for efficient rendering of high‐quality depth‐of‐field (DoF) effects. We start with a single rasterized view (reference view) of the scene, and sample the light field by warping the reference view to nearby views. We implement the algorithm using NVIDIA's CUDA to achieve parallel processing, and exploit the atomic operations to resolve visibility when multiple pixels warp to the same image location. We then directly synthesize DoF effects from the sampled light field. To reduce aliasing artifacts, we propose an image‐space filtering technique that compensates for spatial undersampling using MIP mapping. The main advantages of our algorithm are its simplicity and generality. We demonstrate interactive rendering of DoF effects in several complex scenes. Compared to existing methods, ours does not require ray tracing and hence scales well with scene complexity.

We present a catadioptric projector analogous to a catadioptric camera by combining a commodity digital projector with additional optical units. We show that, by using specially shaped reflectors/refractors, catadioptric projectors can offer an unprecedented level of flexibility in aspect ratio, size, and field of view. We also present efficient algorithms to reduce projection artifacts in catadioptric projectors, such as distortions, scattering, and defocusing. Instead of recovering the reflector/refractor geometry, our approach directly models the light transport between the projector and the viewpoint using the light transport matrix (LTM). We show how to efficiently approximate the pseudo inverse of the LTM and apply it to find the optimal input image that produces least projection distortions. Furthermore, we present a projection defocus analysis for reflector and thin refractor based catadioptric projectors. We show that defocus blur can be interpreted as spatially-varying Gaussian blurs on the input image. We then measure the kernels directly from the LTM and apply deconvolution to optimize the input image. We demonstrate the practical uses of catadioptric projectors in panoramic and omni-directional projections. Our new system achieves much wider field-of-view projection while maintaining sharpness and low geometric and photometric distortions.

We present a new shape-from-distortion framework for recovering specular (reflective/refractive) surfaces. While most existing approaches rely on accurate correspondences between 2D pixels and 3D points, we focus on analyzing the curved images of 3D lines which we call curved line images or CLIs. Our approach models CLIs of local reflections or refractions using the recently proposed general linear cameras (GLCs). We first characterize all possible CLIs in a GLC. We show that a 3D line will appear as a conic in any GLC. For a fixed GLC, the conic type is invariant to the position and orientation of the line and is determined by the GLC parameters. Furthermore, CLIs under single reflection/refraction can only be lines or hyperbolas. Based on our new theory, we develop efficient algorithms to use multiple CLIs to recover the GLC camera parameters. We then apply the curvature-GLC theory to derive the Gaussian and mean curvatures from the GLC intrinsics. This leads to a complete distortion-based reconstruction framework. Unlike conventional correspondence-based approaches that are sensitive to image distortions, our approach benefits from the CLI distortions. Finally, we demonstrate applying our framework for recovering curvature fields on both synthetic and real specular surfaces.

Stereo matching and volumetric reconstruction are the most explored 3D scene recovery techniques in computer vision. Many existing approaches assume perspective input images and use the epipolar constraint to reduce the search space and improve the accuracy. In this paper we present a novel framework that uses multi-perspective cameras for stereo matching and volumetric reconstruction. Our approach first decomposes a multi-perspective camera into piecewise primitive General Linear Cameras or GLCs. A pair of GLCs in general do not satisfy the epipolar constraint. However, they still form a nearly stereo pair. We develop a new Graph-Cut-based algorithm to account for the slight vertical parallax using the GLC ray geometry. We show that the recovered pseudo disparity map conveys important depth cues analogous to perspective stereo matching. To more accurately reconstruct a 3D scene, we develop a new multi-perspective volumetric reconstruction method. We discretize the scene into voxels and apply the GLC back-projections to map the voxel onto each input multi-perspective camera. Finally, we apply the graph-cut algorithm to optimize the 3D embedded voxel graph. We demonstrate our algorithms on both synthetic and real multi-perspective cameras. Experimental results show that our methods are robust and reliable.

In this paper, we propose a method for robustly determining the vignetting function given only a single image. Our method is designed to handle both textured and untextured regions in order to maximize the use of available information. To extract vignetting information from an image, we present adaptations of segmentation techniques that locate image regions with reliable data for vignetting estimation. Within each image region, our method capitalizes on the frequency characteristics and physical properties of vignetting to distinguish it from other sources of intensity variation. Rejection of outlier pixels is applied to improve the robustness of vignetting estimation. Comprehensive experiments demonstrate the effectiveness of this technique on a broad range of images with both simulated and natural vignetting effects. Causes of failures using the proposed algorithm are also analyzed.

We consider the problem of capturing shape characteristics on specular (refractive and reflective) surfaces that are nearly flat. These surfaces are difficult to model using traditional methods based on reconstructing the surface positions and normals. These lower-order shape attributes provide little information to identify important surface characteristics related to distortions. In this paper, we present a framework for recovering the higher-order geometry attributes of specular surfaces. Our method models local reflections and refractions in terms of a special class of multiperspective cameras called the general linear cameras (GLCs). We then develop a new theory that correlates the higher-order differential geometry attributes with the local GLCs. Specifically, we show that Gaussian and mean curvature can be directly derived from the camera intrinsics of the local GLCs. We validate this theory on both synthetic and real-world specular surfaces. Our method places a known pattern in front of a reflective surface or beneath a refractive surface and captures a distorted image on the surface. We then compute the optimal GLC using a sparse set of correspondences and recover the curvatures from the GLC. Experiments demonstrate that our methods are robust and highly accurate.

We present a hybrid camera that combines the advantages of a high resolution camera and a high speed camera. Our hybrid camera consists of a pair of low-resolution high-speed (LRHS) cameras and a single high-resolution low-speed (HRLS) camera. The LRHS cameras are able to capture fast-motion with little motion blur. They also form a stereo pair and provide a low-resolution depth map. The HRLS camera provides a high spatial resolution but also introduces severe motion blur when capturing fast moving objects. We develop efficient algorithms to simultaneously motion-deblur the HRLS image and reconstruct a high resolution depth map. Our method estimates the motion flow in the LRHS pair and then warps the flow field to the HRLS camera to estimate the point spread function (PSF).We then deblur the HRLS image and use the resulting image to enhance the low-resolution depth map using joint bilateral filters. We demonstrate the hybrid camera in depth map super-resolution and motion deblurring with spatially varying kernels. Experiments show that our framework is robust and highly effective.

In this paper, we propose an online interactive matting algorithm, which we call FuzzyMatte. Our framework is based on computing the fuzzy connectedness (FC) [20] from each unknown pixel to the known foreground and background. FC effectively captures the adjacency and similarity between image elements and can be efficiently computed using the strongest connected path searching algorithm. The final alpha value at each pixel can then be calculated from its FC. While many previous methods need to completely recompute the matte when new inputs are provided, FuzzyMatte effectively integrates these new inputs with the previously estimated matte by efficiently recomputing the FC value for a small subset of pixels. Thus, the computational overhead between each iteration of the refinement is significantly reduced. We demonstrate FuzzyMatte on a wide range of images. We show that FuzzyMatte updates the matte in an online interactive setting and generates high quality matte for complex images.

In this paper, we present a novel single-image vignetting method based on the symmetric distribution of the radial gradient (RG). The radial gradient is the image gradient along the radial direction with respect to the image center. We show that the RG distribution for natural images without vignetting is generally symmetric. However, this distribution is skewed by vignetting. We develop two variants of this technique, both of which remove vignetting by minimizing asymmetry of the RG distribution. Compared with prior approaches to single-image vignetting correction, our method does not require segmentation and the results are generally better. Experiments show our technique works for a wide range of images and it achieves a speed-up of 4’5 times compared with a state-of-the-art method.

A perspective image represents the spatial relationships of objects in a scene as they appear from a single viewpoint. In contrast, a multiperspective image combines what is seen from several viewpoints into a single image. Despite their incongruity of view, effective multiperspective images are able to preserve spatial coherence and can depict, within a single context, details of a scene that are simultaneously inaccessible from a single view, yet easily interpretable by a viewer. In computer vision, multiperspective images have been used for analyzing structure revealed via motion and generating panoramic images with a wide field-of-view using mirrors.

Abstract We present an interactive GPU‐based algorithm for accurately rendering high‐quality, dynamic glossy reflection effects from both HDR environment maps and local scene objects. Our method uses hardware rasterization to produce primary pixels, and GPU‐based BRDF importance sampling [ CK07 ] to quickly generate reflected rays. We utilize a fast GPU ray tracer proposed by Carr et al. [ CHCH06 ] to compute reflection hits. Our main contribution is an adaptive level‐of‐detail (LOD) control algorithm that greatly improves ray tracing performance during reflection shading. Specifically, we use the solid angle represented by each reflected ray to adaptively pick the level of termination in the BVH traversal step during ray tracing. This leads to 2 ∼ 3x speedup over an unmodified implementation of [ CHCH06 ]. Based on the same solid angle measure, we derive a texture filtering formula to reduce reflection aliasing artifacts, taking advantage of hardware MIP mapping. This extends the filtering algorithm presented in [ CK07 ] from environment mapping to local scene reflection. Using our algorithm, we demonstrate interactive rendering rates for several scenes featuring dynamic lighting and material changes, spatially varying BRDF parameters, and rigid‐body object movement.

Line-art illustrations are effective tools for conveying shapes and shading of complex objects. We present a set of new algorithms to render line-art illustrations of dynamic and specular (reflective and refractive) surfaces. We first introduce a real-time principal direction estimation algorithm to determine the line stroke directions on dynamic opaque objects using neighboring normal ray triplets. To render reflections or refractions in a line-art style, we develop a stroke direction propagation algorithm by using multi-perspective projections to propagate the stroke directions from the nearby opaque objects onto specular surfaces. Finally, we present an image-space stroke mapping method to draw line strokes using the computed or propagated stroke directions. We implement these algorithms using a GPU and demonstrate real-time illustrations of scenes with dynamic and specular 3D models in line-art styles.

Human stereo vision works by fusing a pair of perspective images with a purely horizontal parallax. Recent developments suggest that very few varieties of multiperspective stereo pairs exist. In this paper, we introduce a new stereo model, which we call epsilon stereo pairs, for fusing a broader class of multiperspective images. An epsilon stereo pair consists of two images with a slight vertical parallax. We show many multiperspective camera pairs that do not satisfy the stereo constraint can still form epsilon stereo pairs. We then introduce a new ray-space warping algorithm to minimize stereo inconsistencies in an epsilon pair using multiperspective collineations. This makes epsilon stereo model a promising tool for synthesizing close-to-stereo fusions from many non-stereo pairs.

In this paper we propose a new framework to simultaneously segment and register lung and tumor in serial CT data. Our method assumes nonrigid transformation on lung deformation and rigid structure on the tumor. We use the B- Spline-based nonrigid transformation to model the lung deformation while imposing rigid transformation on the tumor to preserve the volume and the shape of the tumor. In particular, we set the control points within the tumor to form a control mesh and thus assume the tumor region follows the same rigid transformation as the control mesh. For segmentation, we apply a 2D graph-cut algorithm on the 3D lung and tumor datasets. By iteratively performing segmentation and registration, our method achieves highly accurate segmentation and registration on serial CT data. Finally, since our method eliminates the possible volume variations of the tumor during registration, we can further estimate accurately the tumor growth, an important evidence in lung cancer diagnosis. Initial experiments on five sets of patients ' serial CT data show that our method is robust and reliable.

Caustics are important visual phenomena, as well as challenging global illumination effects in computer graphics. Physically caustics can be interpreted from one of two perspectives: in terms of photons gathered on scene geometry, or in terms of a pair of caustic surfaces. These caustic surfaces are swept by the foci of light rays. In this paper, we develop a novel algorithm to approximate caustic surfaces of sampled rays. Our approach locally parameterizes rays by their intersections with a pair of parallel planes. We show neighboring ray triplets are constrained to pass simultaneously through two slits, which rule the caustic surfaces. We derive a ray characteristic equation to compute the two slits, and hence, the caustic surfaces. Using the characteristic equation, we develop a GPU-based algorithm to render the caustics. Our approach produces sharp and clear caustics using much fewer ray samples than the photon mapping method and it also maintains high spatial and temporal coherency. Finally, we present a normal-ray surface representation that locally parameterizes the normals about a surface point as rays. Computing the normal ray caustic surfaces leads to a novel real-time discrete shape operator.

Illumination inconsistencies cause serious problems for classical computer vision applications such as tracking and stereo matching. We present a new approach to model illumination variations using an Illumination Ratio Map (IRM). An IRM computes the intensity ratio of corresponding points in an image pair. We formulate IRM recovery as a Markov network, which assumes spatially varying illumination changes can be modeled as a locally smooth function with boundaries. We show that the IRM Markov network can be easily incorporated into low-level vision problems, such as tracking and stereo matching, by integrating IRM estimation with the optical flow field/disparity map solution process. This leads to a unified Markov network. We develop an iterative optimization algorithm based on Belief Propagation to efficiently recover the illumination ratio map and the optical field/disparity map at the same time. Experiments demonstrate that our methods are robust and reliable.

We present a novel method for analyzing reflections on arbitrary surfaces. We model reflections using a broader than usual class of imaging models, which include both perspective and multiperspective camera types. We provide an analytical framework to locally model reflections as specific multiperspective cameras around every ray based on a new theory of general linear cameras. Our framework better characterizes the complicated image distortions seen on irregular mirror surfaces as well as the conventional catadioptric mirrors. We show the connection between multiperspective camera models and caustic surfaces of reflections and demonstrate how they reveal important surface rulings of the caustics. Finally, we show how to use our analysis to assist mirror design and characterize distortions seen in catadioptric imaging systems.

We present theories of multiperspective projection and collineation. Given an arbitrary multiperspective imaging system that captures smoothly varying set of rays, we show how to map the rays onto a 2D ray manifold embedded in a 4D linear vector space. The characteristics of this imaging system, such as its projection, collineation, and image distortions can be analyzed by studying the 2D tangent planes of this ray manifold. These tangent planes correspond to the recently proposed general linear camera (GLC) model. In this paper, we study the imaging process of the GLCs. We show the GLC imaging process can be broken down into two separate stages: the mapping of 3D geometry to rays and the sampling of those rays over an image plane. We derive a closed-form solution to projecting 3D points in a scene to rays in a GLC. A GLC image is created by sampling these rays over an image plane. We develop a notion of GLC collineation analogous to pinhole cameras. GLC collineation describes the transformation between the images of a single GLC due to changes in sampling and image plane selection. We show that general GLC collineations can be characterized by a quartic (4th order) rational function. GLC projection and collineation provides a basis for developing new computer vision algorithms suitable for analyzing a wider range of imaging systems than current methods, based on simple pinhole projection models, permit.

We present a novel algorithm to efficiently render accurate reflections on programmable graphics hardware. Our algorithm overcomes problems that commonly occur in environment mapping such as the lack of motion parallax and inaccuracies when objects are close to the reflectors. In place of a 2D environment map, which only represents points infinitely far away from the reflector, we use six 4D light field slabs to represent the surrounding scene. Each reflected ray is rendered by indexing into these precaptured environment light fields. We are able to render accurate reflections with motion parallax at interactive frame rates independent of the reflector geometry and the scene complexity. Furthermore, we can move the reflectors within a constrained region of space and guarantee that the environment light field provides the necessary rays. We benefit from the programmability of existing graphics hardware to efficiently compute the reflected rays and transform them into the appropriate light field index. We also take advantage of the large texture memories and memory bandwidth available in today's graphics card to store and query hardware-compressed light fields.

We propose a class of enhancement techniques suitable for scenes captured by fixed cameras. The basic idea is to increase the information density in a set of low quality images by extracting the context from a higher-quality image captured under different illuminations from the same viewpoint. For example, a night-time surveillance video can be enriched with information available in daytime images. We also propose a new image fusion approach to combine images with sufficiently different appearance into a seamless rendering. Our method ensures the fidelity of important features and robustly incorporates background contexts, while avoiding traditional problems such as aliasing, ghosting and haloing. We show results on indoor as well as outdoor scenes.

We present a non-photorealistic rendering approach to capture and convey shape features of real-world scenes. We use a camera with multiple flashes that are strategically positioned to cast shadows along depth discontinuities in the scene. The projective-geometric relationship of the camera-flash setup is then exploited to detect depth discontinuities and distinguish them from intensity edges due to material discontinuities.We introduce depiction methods that utilize the detected edge features to generate stylized static and animated images. We can highlight the detected features, suppress unnecessary details or combine features from multiple images. The resulting images more clearly convey the 3D structure of the imaged scenes.We take a very different approach to capturing geometric features of a scene than traditional approaches that require reconstructing a 3D model. This results in a method that is both surprisingly simple and computationally efficient. The entire hardware/software setup can conceivably be packaged into a self-contained device no larger than existing digital cameras.

In this paper we present a new variant of the light field representation that supports improved image reconstruction by accommodating sparse correspondence information. This places our representation somewhere between a pure, two-plane parameterized, light field and a lumigraph representation, with its continuous geometric proxy. Our approach factorises the rays of a light field into one of two separate classes. All rays consistent with a given correspondence are implicitly represented using a new auxiliary data structure, which we call a surface camera (or scam). The remaining rays of the light field are represented using a standard two-plane parameterized light field. We present an efficient rendering algorithm that combines ray samples from scams with those from the light field. The resulting image reconstructions are noticeably improved over that of a pure light field.