Automating immersive VR scene creation remains a primary research challenge. Existing methods typically rely on complex geometry with post-simplification, resulting in inefficient pi pelines or li mited re alism. In th is paper, we introduce Im merseGen, a novel agent-guided framework for compact and photorealistic world generation that decouples realism from exhaustive geometric modeling. ImmerseGen represents scenes as hierarchical compositions of lightweight geometric proxies with synthesized RGBA textures, facilitating real-time rendering on mobile VR headsets. We propose terrain-conditioned texturing for base world generation, combined with context-aware texturing for scenery, to produce diverse and visually coherent worlds. VLM-based agents employ semantic grid-based analysis for precise asset placement and enrich scenes with multimodal enhancements such as visual dynamics and ambient sound. Experiments and real-time VR applications demonstrate that ImmerseGen achieves superior photorealism, spatial coherence, and rendering efficiency compared to existing methods.

Neural surface representation has demonstrated remarkable success in the areas of novel view synthesis and 3D reconstruction. However, assessing the geometric quality of 3D reconstructions in the absence of ground truth mesh remains a significant challenge, due to its rendering-based optimization process and entangled learning of appearance and geometry with photometric losses. In this paper, we present a novel framework, i.e, GURecon, which establishes a geometric uncertainty field for the neural surface based on geometric consistency. Different from existing methods that rely on rendering-based measurement, GURecon models a continuous 3D uncertainty field for the reconstructed surface, and is learned by an online distillation approach without introducing real geometric information for supervision. Moreover, in order to mitigate the interference of illumination on geometric consistency, a decoupled field is learned and exploited to finetune the uncertainty field. Experiments on various datasets demonstrate the superiority of GURecon in modeling 3D geometric uncertainty, as well as its plug-and-play extension to various neural surface representations and improvement on downstream tasks such as incremental reconstruction.

Recent advances in deep learning have significantly boosted performance in multivariate time series forecasting (MTSF). While many existing approaches focus on capturing inter-variable (a.k.a. channel-wise) correlations to improve prediction accuracy, the temporal dimension, particularly its rich structural and contextual information, remains underexplored. In this paper, we propose BIM3, a novel framework that integrates BIdirectional temporal-aware modeling with Multi-Scale Mixture-of-Experts for MTSF. First, unlike existing methods that treat historical and future temporal information independently, we introduce a novel Timestamp Dual Cross-Attention Module, which employs a symmetric cross-attention mechanism to explicitly capture bidirectional temporal dependencies through timestamp interactions. Second, to address the complex and scale-varying temporal patterns commonly found in multivariate time series, we move beyond recent multi-scale forecasting models that share parameters across all channels and fail to capture channel-specific dynamics. Instead, we design a Multi-Scale Feature Extract Mixture-of-Experts module that adaptively routes time series to specialized experts based on their temporal characteristics. Extensive experiments on multiple real-world datasets show that BIM3 consistently outperforms state-of-the-art methods, highlighting its effectiveness in capturing both temporal structure and inter-variable diversity.

Neural rendering has demonstrated remarkable success in high-quality 3D neural reconstruction and novel view synthesis with dense input views and accurate poses. However, applying it to extremely sparse, unposed views in unbounded 360° scenes remains a challenging problem. In this paper, we propose a novel neural rendering framework to accomplish the unposed and extremely sparse-view 3D reconstruction in unbounded 360° scenes. To resolve the spatial ambiguity inherent in unbounded scenes with sparse input views, we propose a layered Gaussian-based representation to effectively model the scene with distinct spatial layers. By employing a dense stereo reconstruction model to recover coarse geometry, we introduce a layer-specific bootstrap optimization to refine the noise and fill occluded regions in the reconstruction. Furthermore, we propose an iterative fusion of reconstruction and generation alongside an uncertainty-aware training approach to facilitate mutual conditioning and enhancement between these two processes. Comprehensive experiments show that our approach outperforms existing state-of-the-art methods in terms of rendering quality and surface reconstruction accuracy. Project page: https://zju3dv.github.io/free360/.

We propose a novel 3D-aware diffusion-based method for generating photorealistic talking head videos directly from a single identity image and explicit control signals (e.g., expressions). Our method generates Multiplane Images (MPIs) that ensure geometric consistency, making them ideal for immersive viewing experiences like binocular videos for VR headsets. Unlike existing methods that often require a separate stage or joint optimization to reconstruct a 3D representation (such as NeRF or 3D Gaussians), our approach directly generates the final output through a single denoising process, eliminating the need for post-processing steps to render novel views efficiently. To effectively learn from monocular videos, we introduce a training mechanism that reconstructs the output MPI randomly in either the target or the reference camera space. This approach enables the model to simultaneously learn sharp image details and underlying 3D information. Extensive experiments demonstrate the effectiveness of our method, which achieves competitive avatar quality and novel-view rendering capabilities, even without explicit 3D reconstruction or high-quality multi-view training data.

Event cameras, known for their high temporal resolution and ability to capture asynchronous changes, have gained significant attention for their potential in feature tracking, especially in challenging conditions. However, event cameras lack the fine-grained texture information that conventional cameras provide, leading to error accumulation in tracking. To address this, we propose a novel framework, BlinkTrack, which integrates event data with grayscale images for high-frequency feature tracking. Our method extends the traditional Kalman filter into a learning-based framework, utilizing differentiable Kalman filters in both event and image branches. This approach improves single-modality tracking and effectively solves the data association and fusion from asynchronous event and image data. We also introduce new synthetic and augmented datasets to better evaluate our model. Experimental results indicate that BlinkTrack significantly outperforms existing methods, exceeding 80 FPS with multi-modality data and 100 FPS with preprocessed event data. Codes and dataset are available at https://github.com/ColieShen/BlinkTrack.

Humans can naturally identify and mentally complete occluded objects in cluttered environments. However, imparting similar cognitive ability to robotics remains challenging even with advanced reconstruction techniques, which models scenes as undifferentiated wholes and fails to recognize complete object from partial observations. In this paper, we propose InstaScene, a new paradigm towards holistic 3D perception of complex scenes with a primary goal: decomposing arbitrary instances while ensuring complete reconstruction. To achieve precise decomposition, we develop a novel spatial contrastive learning by tracing rasterization of each instance across views, significantly enhancing semantic supervision in cluttered scenes. To overcome incompleteness from limited observations, we introduce in-situ generation that harnesses valuable observations and geometric cues, effectively guiding 3D generative models to reconstruct complete instances that seamlessly align with the real world. Experiments on scene decomposition and object completion across complex real-world and synthetic scenes demonstrate that our method achieves superior decomposition accuracy while producing geometrically faithful and visually intact objects.

Recently, neural radiance fields (NeRF) have gained significant attention in the field of visual localization. However, existing NeRF-based approaches either lack geometric constraints or require extensive storage for feature matching, limiting their practical applications. To address these challenges, we propose an efficient and novel visual localization approach based on the neural implicit map with complementary features. Specifically, to enforce geometric constraints and reduce storage requirements, we implicitly learn a 3D keypoint descriptor field, avoiding the need to explicitly store point-wise features. To further address the semantic ambiguity of descriptors, we introduce additional semantic contextual feature fields, which enhance the quality and reliability of 2D-3D correspondences. Besides, we propose descriptor similarity distribution alignment to minimize the domain gap between 2D and 3D feature spaces during matching. Finally, we construct the matching graph using both complementary descriptors and contextual features to establish accurate 2D3D correspondences for 6-DoF pose estimation. Compared with the recent NeRF-based approaches, our method achieves a$3 \times$faster training speed and a$45 \times$reduction in model storage. Extensive experiments on two widely used datasets demonstrate that our approach outperforms or is highly competitive with other state-of-the-art NeRF-based visual localization methods. Project page: https://zju3dv.github.io/neuraloc

Visual-inertial odometry (VIO) has made significant progress in various applications. However, one of the key challenges in VIO is the efficient and robust fusion of visual and inertial measurements, particularly while mitigating the impact of sensor failures. To address this challenge, we propose a new learning-based VIO system, i.e., DW-VIO, which is able to integrate multiple sensors and provide robust state estimations. To this end, we design a novel deep learning-based data-fusion approach that dynamically associates information from multiple sensors to predict sensor weights for optimization. Moreover, in order to improve the efficiency, we present several real-time optimization techniques including a fast patch graph constructor and an efficient GPU-accelerated multi-factor bundle adjustment layer. Experimental results show that DW-VIO outperforms most state-of-the-art (SOTA) methods on the EuRoC MAV, ETH3D-SLAM, and KITTI-360 benchmarks across various challenging sequences. Additionally, it maintains a minimum of 20 frames per second (FPS) on a single RTX 3060 GPU with high-resolution input, highlighting its efficiency.

Human video synthesis aims to create lifelike characters in various environments. While 2D diffusion-based methods have made significant progress, they struggle to generalize to complex 3D poses and varying scene backgrounds. To address these limitations, we introduce CFSynthesis, a novel framework for generating high-quality human videos with customizable attributes, including identity, motion, and scene configurations. Our method leverages a texture-SMPL-based representation to ensure consistent and stable character appearances across free viewpoints. Additionally, we introduce a novel foreground-background separation strategy that effectively decomposes the scene as foreground and background, enabling seamless integration of user-defined backgrounds. Experimental results on multiple datasets show that CFSynthesis not only achieves state-of-the-art performance in complex human animations but also adapts effectively to 3D motions in free-view and user-specified scenarios.

Scene-level 3D generation represents a critical frontier in multimedia and computer graphics. While existing approaches have achieved encouraging progress, they still face challenges such as constrained object diversity and limited support for interactive editing. In this paper, we present HiScene, a novel hierarchical framework that bridges the gap between 2D image generation and 3D object generation and delivers high-fidelity scenes with compositional identities and aesthetic scene content. Our key insight is treating scenes as hierarchical ''objects'' under isometric views, where a room functions as a complex object that can be further decomposed into manipulatable items. This hierarchical approach enables us to generate 3D content that aligns with 2D representations while maintaining compositional structure. To ensure completeness and spatial alignment of each decomposed instance, we develop a video-diffusion-based amodal completion technique that effectively handles occlusions and shadows between objects, and introduce shape prior injection to ensure spatial coherence within the scene. Experimental results demonstrate that our method produces more natural object arrangements and complete object instances suitable for interactive applications, while maintaining physical plausibility and alignment with user inputs.

3D reconstruction of indoor and urban environments is a prominent research topic with various downstream applications. However, existing geometric priors for addressing low-texture regions in indoor and urban settings often lack global consistency. Moreover, Gaussian Splatting and implicit SDF fields often suffer from discontinuities or exhibit computational inefficiencies, resulting in a loss of detail. To address these issues, we propose an Atlanta-world guided implicit-structured Gaussian Splatting that achieves smooth indoor and urban scene reconstruction while preserving high-frequency details and rendering efficiency. By leveraging the Atlanta-world model, we ensure the accurate surface reconstruction for low-texture regions, while the proposed novel implicit-structured GS representations provide smoothness without sacrificing efficiency and high-frequency details. Specifically, we propose a semantic GS representation to predict the probability of all semantic regions and deploy a structure plane regularization with learnable plane indicators for global accurate surface reconstruction. Extensive experiments demonstrate that our method outperforms state-of-the-art approaches in both indoor and urban scenes, delivering superior surface reconstruction quality.

In this paper, we present TexPro, a novel method for high-fidelity material generation for input 3D meshes given text prompts. Unlike existing text-conditioned texture generation methods that typically generate RGB textures with baked lighting, TexPro is able to produce diverse texture maps via procedural material modeling, which enables physically-based rendering, relighting, and additional benefits inherent to procedural materials. Specifically, we first generate multi-view reference images given the input textual prompt by employing the latest text-to-image model. We then derive texture maps through rendering-based optimization with recent differentiable procedural materials. To this end, we design several techniques to handle the misalignment between the generated multiview images and 3D meshes, and introduce a novel material agent that enhances material classification and matching by exploring both part-level understanding and object-aware material reasoning. Experiments demonstrate the superiority of the proposed method over existing SOTAs, and its capability of relighting.

Recently neural implicit rendering techniques have evolved rapidly and demonstrated significant advantages in novel view synthesis and 3D scene reconstruction. However, existing neural rendering methods for editing purposes offer limited functionalities, e.g., rigid transformation and category-specific editing. In this paper, we present a novel mesh-based representation by encoding the neural radiance field with disentangled geometry, texture, and semantic codes on mesh vertices, which empowers a set of efficient and comprehensive editing functionalities, including mesh-guided geometry editing, designated texture editing with texture swapping, filling and painting operations, and semantic-guided editing. To this end, we develop several techniques including a novel local space parameterization to enhance rendering quality and training stability, a learnable modification color on vertex to improve the fidelity of texture editing, a spatial-aware optimization strategy to realize precise texture editing, and a semantic-aided region selection to ease the laborious annotation of implicit field editing. Extensive experiments and editing examples on both real and synthetic datasets demonstrate the superiority of our method on representation quality and editing ability.

Neural radiance field (NeRF) has achieved impressive results in high-quality 3D scene reconstruction. However, NeRF heavily relies on precise camera poses. While recent works like BARF have introduced camera pose optimization within NeRF, their applicability is limited to simple trajectory scenes. Existing methods struggle while tackling complex trajectories involving large rotations. To address this limitation, we propose CT-NeRF, an incremental reconstruction and optimization pipeline using only RGB images without pose and depth input. In this pipeline, we first propose a local-global bundle adjustment under a pose graph connecting neighboring frames to enforce the consistency between poses to escape the local minima caused by only pose consistency with the scene structure. Further, we instantiate the consistency between poses as a reprojection error constraint resulting from pixel-level correspondences between input image pairs. Through the incremental reconstruction, CT-NeRF enables the recovery of both camera poses and scene structure and is capable of handling scenes with complex trajectories. We evaluate the performance of CT-NeRF on two real-world datasets, NeRF-Buster and Free-Dataset, which feature complex trajectories. Results show CT-NeRF outperforms existing methods in novel view synthesis and pose estimation accuracy.

Visual localization plays an important role in the applications of Augmented Reality (AR), which enable AR devices to obtain their 6-DoF pose in the pre-build map in order to render virtual content in real scenes. However, most existing approaches can not perform novel view rendering and require large storage capacities for maps. To overcome these limitations, we propose an efficient visual localization method capable of high-quality rendering with fewer parameters. Specifically, our approach leverages 3D Gaussian primitives as the scene representation. To ensure precise 2D-3D correspondences for pose estimation, we develop an unbiased 3D scene-specific descriptor decoder for Gaussian primitives, distilled from a constructed feature volume. Additionally, we introduce a salient 3D landmark selection algorithm that selects a suitable primitive subset based on the saliency score for localization. We further regularize key Gaussian primitives to prevent anisotropic effects, which also improves localization performance. Extensive experiments on two widely used datasets demonstrate that our method achieves superior or comparable rendering and localization performance to state-of-the-art implicit-based visual localization approaches. Code and data are available at project page: https://zju3dv.github.io/splatloc.

Neural implicit representations have recently become popular in simultaneous localization and mapping (SLAM), especially in dense visual SLAM. However, existing works either rely on RGB-D sensors or require a separate monocular SLAM approach for camera tracking, and fail to produce high-fidelity 3D dense reconstructions. To address these shortcomings, we present NICER-SLAM, a dense RGB SLAM system that simultaneously optimizes for camera poses and a hierarchical neural implicit map representation, which also allows for high-quality novel view synthesis. To facilitate the optimization process for mapping, we integrate additional supervision signals including easy-to-obtain monocular geometric cues and optical flow, and also introduce a simple warping loss to further enforce geometric consistency. Moreover, to further boost performance in complex large-scale scenes, we also propose a local adaptive transformation from signed distance functions (SDFs) to density in the volume rendering equation. On multiple challenging indoor and outdoor datasets, NICER-SLAM demonstrates strong performance in dense mapping, novel view synthesis, and tracking, even competitive with recent RGB-D SLAM systems. Project page: https://nicer-slam.github.io/.

Due to the ability to synthesize high-quality novel views, Neural Radiance Fields (NeRF) has been recently exploited to improve visual localization in a known environment. However, the existing methods mostly utilize NeRF for data augmentation to improve the regression model training, and their performances on novel viewpoints and appearances are still limited due to the lack of geometric constraints. In this paper, we propose a novel visual localization framework, i.e., PNeRFLoc, based on a unified point-based representation. On one hand, PNeRFLoc supports the initial pose estimation by matching 2D and 3D feature points as traditional structure-based methods; on the other hand, it also enables pose refinement with novel view synthesis using rendering-based optimization. Specifically, we propose a novel feature adaption module to close the gaps between the features for visual localization and neural rendering. To improve the efficacy and efficiency of neural rendering-based optimization, we also developed an efficient rendering-based framework with a warping loss function. Extensive experiments demonstrate that PNeRFLoc performs the best on the synthetic dataset when the 3D NeRF model can be well learned, and significantly outperforms all the NeRF-boosted localization methods with on-par SOTA performance on the real-world benchmark localization datasets. Project webpage: https://zju3dv.github.io/PNeRFLoc/.

Recently, we have witnessed the explosive growth of various volumetric representations in modeling animatable head avatars. However, due to the diversity of frameworks, there is no practical method to support high-level applications like 3D head avatar editing across different representations. In this paper, we propose a generic avatar editing approach that can be universally applied to various 3DMM-driving volumetric head avatars. To achieve this goal, we design a novel expression-aware modification generative model, which enables lift 2D editing from a single image to a consistent 3D modification field. To ensure the effectiveness of the generative modification process, we develop several techniques, including an expression-dependent modification distillation scheme to draw knowledge from the large-scale head avatar model and 2D facial texture editing tools, implicit latent space guidance to enhance model convergence, and a segmentation-based loss reweight strategy for fine-grained texture inversion. Extensive experiments demonstrate that our method delivers high-quality and consistent results across multiple expression and viewpoints. Project page: https://zju3dv.github.io/geneavatar/.

Directly generating scenes from satellite imagery offers exciting possibilities for integration into applications like games and map services. However, challenges arise from significant view changes and scene scale. Previous efforts mainly focused on image or video generation, lacking exploration into the adaptability of scene generation for ar-bitrary views. Existing 3D generation works either oper-ate at the object level or are difficult to utilize the geometry obtained from satellite imagery. To overcome these limitations, we propose a novel architecture for direct 3D scene generation by introducing diffusion models into 3D sparse representations and combining them with neural rendering techniques. Specifically, our approach generates texture colors at the point level for a given geometry using a 3D diffusion model first, which is then transformed into a scene representation in a feed-forward manner. The representation can be utilized to render arbitrary views which would excel in both single-frame quality and inter-frame consistency. Experiments in two city-scale datasets show that our model demonstrates proficiency in generating photorealistic street-view image sequences and cross-view urban scenes from satellite imagery.

In this paper, we propose a novel multi-view stereo (MVS) framework that gets rid of the depth range prior. Unlike recent prior-free MVS methods that work in a pair-wise manner, our method simultaneously considers all the source images. Specifically, we introduce a Multi-view Disparity Attention (MDA) module to aggregate long-range context information within and across multi-view images. Considering the asymmetry of the epipolar disparity flow, the key to our method lies in accurately modeling multi-view geometric constraints. We integrate pose embedding to encapsulate information such as multi-view camera poses, providing implicit geometric constraints for multi-view disparity feature fusion dominated by attention. Additionally, we construct corresponding hidden states for each source image due to significant differences in the observation quality of the same pixel in the reference frame across multiple source frames. We explicitly estimate the quality of the current pixel corresponding to sampled points on the epipolar line of the source image and dynamically update hidden states through the uncertainty estimation module. Extensive results on the DTU dataset and Tanks\&Temple benchmark demonstrate the effectiveness of our method.

Motion Planning (MP) is a critical challenge in robotics, especially pertinent with the burgeoning interest in embodied artificial intelligence. Traditional MP methods often struggle with high-dimensional complexities. Recently neural motion planners, particularly physics-informed neural planners based on the Eikonal equation, have been proposed to overcome the curse of dimensionality. However, these methods perform poorly in complex scenarios with shaped robots due to multiple solutions inherent in the Eikonal equation. To address these issues, this paper presents PC-Planner, a novel physics-constrained self-supervised learning framework for robot motion planning with various shapes in complex environments. To this end, we propose several physical constraints, including monotonic and optimal constraints, to stabilize the training process of the neural network with the Eikonal equation. Additionally, we introduce a novel shape-aware distance field that considers the robot's shape for efficient collision checking and Ground Truth (GT) speed computation. This field reduces the computational intensity, and facilitates adaptive motion planning at test time. Experiments in diverse scenarios with different robots demonstrate the superiority of the proposed method in efficiency and robustness for robot motion planning, particularly in complex environments.

Diffusion-based methods have achieved prominent success in generating 2D media. However, accomplishing similar proficiencies for scene-level mesh texturing in 3D spatial applications, e.g., XR/VR, remains constrained, primarily due to the intricate nature of 3D geometry and the necessity for immersive free-viewpoint rendering. In this paper, we propose a novel indoor scene texturing framework, which delivers text-driven texture generation with enchanting details and authentic spatial coherence. The key insight is to first imagine a stylized 360° panoramic texture from the central viewpoint of the scene, and then propagate it to the rest areas with inpainting and imitating techniques. To ensure meaningful and aligned textures to the scene, we develop a novel coarse-to-fine panoramic texture generation approach with dual texture alignment, which both considers the geometry and texture cues of the captured scenes. To survive cluttered geometries during texture propagation, we design a separated strategy, which conducts texture inpainting in visible regions and then learns an implicit imitating network to synthesize textures in occluded and tiny structural areas. Extensive experiments and the immersive VR application on real-world indoor scenes demonstrate the high quality of the generated textures and the engaging experience on VR headsets. Project webpage: https://ybbbbt.com/publication/dreamspace.

Structure-from-Motion (SfM) aims to recover 3D scene structures and camera poses based on the correspondences between input images, and thus the ambiguity caused by duplicate structures (i.e., different structures with strong visual resemblance) always results in incorrect camera poses and 3D structures. To deal with the ambiguity, most existing studies resort to additional constraint information or implicit inference by analyzing two-view geometries or feature points. In this paper, we propose to exploit high-level information in the scene, i.e., the spatial contextual information of local regions, to guide the reconstruction. Specifically, a novel structure is proposed, namely, track-community, in which each community consists of a group of tracks and represents a local segment in the scene. A community detection algorithm is performed on the track-graph to partition the scene into segments. Then, the potential ambiguous segments are detected by analyzing the neighborhood of tracks and corrected by checking the pose consistency. Finally, we perform partial reconstruction on each segment and align them with a novel bidirectional consistency cost function which considers both 3D-3D correspondences and pairwise relative camera poses. Experimental results demonstrate that our approach can robustly alleviate reconstruction failure resulting from visually indistinguishable structures and accurately merge the partial reconstructions.

Despite the great success in 2D editing using user-friendly tools, such as Photoshop, semantic strokes, or even text prompts, similar capabilities in 3D areas are still limited, either relying on 3D modeling skills or allowing editing within only a few categories. In this paper, we present a novel semantic-driven NeRF editing approach, which enables users to edit a neural radiance field with a single image, and faithfully delivers edited novel views with high fidelity and multi-view consistency. To achieve this goal, we propose a prior-guided editing field to encode fine-grained geometric and texture editing in 3D space, and develop a series of techniques to aid the editing process, including cyclic constraints with a proxy mesh to facilitate geometric supervision, a color compositing mechanism to stabilize semantic-driven texture editing, and a feature-cluster-based regularization to preserve the irrelevant content unchanged. Extensive experiments and editing examples on both real-world and synthetic data demonstrate that our method achieves photo-realistic 3D editing using only a single edited image, pushing the bound of semantic-driven editing in 3D real-world scenes.

Local feature matching aims at establishing sparse correspondences between a pair of images. Recently, detector-free methods present generally better performance but are not satisfactory in image pairs with large scale differences. In this paper, we propose Patch Area Transportation with Subdivision (PATS) to tackle this issue. Instead of building an expensive image pyramid, we start by splitting the original image pair into equal-sized patches and gradually resizing and subdividing them into smaller patches with the same scale. However, estimating scale differences between these patches is non-trivial since the scale differences are determined by both relative camera poses and scene structures, and thus spatially varying over image pairs. Moreover, it is hard to obtain the ground truth for real scenes. To this end, we propose patch area transportation, which enables learning scale differences in a self-supervised manner. In contrast to bipartite graph matching, which only handles one-to-one matching, our patch area transportation can deal with many-to-many relationships. PATS improves both matching accuracy and coverage, and shows superior performance in downstream tasks, such as relative pose estimation, visual localization, and optical flow estimation. The source code is available at https://zju3dv.github.io/pats/.

We present PVO, a novel panoptic visual odometry framework to achieve more comprehensive modeling of the scene motion, geometry, and panoptic segmentation information. Our PVO models visual odometry (VO) and video panoptic segmentation (VPS) in a unified view, which makes the two tasks mutually beneficial. Specifically, we introduce a panoptic update module into the VO Module with the guidance of image panoptic segmentation. This Panoptic-Enhanced VO Module can alleviate the impact of dynamic objects in the camera pose estimation with a panoptic-aware dynamic mask. On the other hand, the VO-Enhanced VPS Module also improves the segmentation accuracy by fusing the panoptic segmentation result of the current frame on the fly to the adjacent frames, using geometric information such as camera pose, depth, and optical flow obtained from the VO Module. These two modules contribute to each other through recurrent iterative optimization. Extensive experiments demonstrate that PVO outperforms state-of-the-art methods in both visual odometry and video panoptic segmentation tasks.

Light-weight time-of-flight (ToF) depth sensors are compact and cost-efficient, and thus widely used on mobile devices for tasks such as autofocus and obstacle detection. However, due to the sparse and noisy depth measurements, these sensors have rarely been considered for dense geometry reconstruction. In this work, we present the first dense SLAM system with a monocular camera and a light-weight ToF sensor. Specifically, we propose a multi-modal implicit scene representation that supports rendering both the signals from the RGB camera and light-weight ToF sensor which drives the optimization by comparing with the raw sensor inputs. Moreover, in order to guarantee successful pose tracking and reconstruction, we exploit a predicted depth as an intermediate supervision and develop a coarse-to-fine optimization strategy for efficient learning of the implicit representation. At last, the temporal information is explicitly exploited to deal with the noisy signals from light-weight ToF sensors to improve the accuracy and robustness of the system. Experiments demonstrate that our system well exploits the signals of light-weight ToF sensors and achieves competitive results both on camera tracking and dense scene reconstruction. Project page: https://zju3dv.github.io/tof_slam/.

Hand-object interaction understanding and the barely addressed novel view synthesis are highly desired in the immersive communication, whereas it is challenging due to the high deformation of hand and heavy occlusions between hand and object. In this paper, we propose a neural rendering and pose estimation system for hand-object interaction from sparse views, which can also enable 3D hand-object interaction editing. We share the inspiration from recent scene understanding work that shows a scene specific model built beforehand can significantly improve and unblock vision tasks especially when inputs are sparse, and extend it to the dynamic hand-object interaction scenario and propose to solve the problem in two stages. We first learn the shape and appearance prior knowledge of hands and objects separately with the neural representation at the offline stage. During the online stage, we design a rendering-based joint model fitting framework to understand the dynamic hand-object interaction with the pre-built hand and object models as well as interaction priors, which thereby overcomes penetration and separation issues between hand and object and also enables novel view synthesis. In order to get stable contact during the hand-object interaction process in a sequence, we propose a stable contact loss to make the contact region to be consistent. Experiments demonstrate that our method outperforms the state-of-the-art methods. Code and dataset are available in project web-page https://iscas3dv.github.io/HO-NeRF.

Stereo depth estimation usually struggles to deal with textureless scenes for both traditional and learning-based methods due to the inherent dependence on image correspondence matching. In this paper, we propose a novel neural network, i.e., DPS-Net, to exploit both the prior geometric knowledge and polarimetric information for depth estimation with two polarimetric stereo images. Specifically, we construct both RGB and polarization correlation volumes to fully leverage the multi-domain similarity between polarimetric stereo images. Since inherent ambiguities exist in the polarization images, we introduce the iso-depth cost explicitly into the network to solve these ambiguities. Moreover, we design a cascaded dual-GRU architecture to recurrently update the disparity and effectively fuse both the multi-domain correlation features and the iso-depth cost. Besides, we present new synthetic and real polarimetric stereo datasets for evaluation. Experimental results demonstrate that our method outperforms the state-of-the-art stereo depth estimation methods.

Existing inverse rendering combined with neural rendering methods can only perform editable novel view synthesis on object-specific scenes, while we present intrinsic neural radiance fields, dubbed IntrinsicNeRF, which introduce intrinsic decomposition into the NeRF-based neural rendering method and can extend its application to room-scale scenes. Since intrinsic decomposition is a fundamentally under-constrained inverse problem, we propose a novel distance-aware point sampling and adaptive reflectance iterative clustering optimization method, which enables IntrinsicNeRF with traditional intrinsic decomposition constraints to be trained in an unsupervised manner, resulting in multi-view consistent intrinsic decomposition results. To cope with the problem that different adjacent instances of similar reflectance in a scene are incorrectly clustered together, we further propose a hierarchical clustering method with coarse-to-fine optimization to obtain a fast hierarchical indexing representation. It supports compelling real-time augmented applications such as recoloring and illumination variation. Extensive experiments and editing samples on both object-specific/room-scale scenes and synthetic/real-word data demonstrate that we can obtain consistent intrinsic decomposition results and high-fidelity novel view synthesis even for challenging sequences.

Event cameras provide high temporal precision, low data rates, and high dynamic range visual perception, which are well-suited for optical flow estimation. While data-driven optical flow estimation has obtained great success in RGB cameras, its generalization performance is seriously hindered in event cameras mainly due to the limited and biased training data. In this paper, we present a novel simulator, BlinkSim, for the fast generation of large-scale data for event-based optical flow. BlinkSim incorporates a configurable rendering engine alongside an event simulation suite. By leveraging the wealth of current 3D assets, the rendering engine enables us to automatically build up thousands of scenes with different objects, textures, and motion patterns and render very high-frequency images for realistic event data simulation. Based on BlinkSim, we construct a large training dataset and evaluation benchmark BlinkFlow that contains sufficient, diversiform, and challenging event data with optical flow ground truth. Experiments show that BlinkFlow improves the generalization performance of state-of-the-art methods by more than 40% on average and up to 90%. Moreover, we further propose an Event-based optical Flow transFormer (E-FlowFormer) architecture. Powered by our BlinkFlow, E-FlowFormer outperforms the SOTA methods by up to 91% on the MVSEC dataset and 14% on the DSEC dataset and presents the best generalization performance. The source code and data are available at https://zju3dv.github.io/blinkflow/.

Recently, Neural Radiance Fields (NeRF) has exhibited significant success in novel view synthesis, surface reconstruction, etc. However, since no physical reflection is considered in its rendering pipeline, NeRF mistakes the reflection in the mirror as a separate virtual scene, leading to the inaccurate reconstruction of the mirror and multi-view inconsistent reflections in the mirror. In this paper, we present a novel neural rendering framework, named Mirror-NeRF, which is able to learn accurate geometry and reflection of the mirror and support various scene manipulation applications with mirrors, such as adding new objects or mirrors into the scene and synthesizing the reflections of these new objects in mirrors, controlling mirror roughness, etc. To achieve this goal, we propose a unified radiance field by introducing the reflection probability and tracing rays following the light transport model of Whitted Ray Tracing, and also develop several techniques to facilitate the learning process. Experiments and comparisons on both synthetic and real datasets demonstrate the superiority of our method. The code and supplementary material are available on the project webpage: https://zju3dv.github.io/Mirror-NeRF/.

This paper presents a collaborative implicit neural simultaneous localization and mapping (SLAM) system with RGB-D image sequences, which consists of complete front-end and back-end modules including odometry, loop detection, sub-map fusion, and global refinement. In order to enable all these modules in a unified framework, we propose a novel neural point based 3D scene representation in which each point maintains a learnable neural feature for scene encoding and is associated with a certain keyframe. Moreover, a distributed-to-centralized learning strategy is proposed for the collaborative implicit SLAM to improve consistency and cooperation. A novel global optimization framework is also proposed to improve the system accuracy like traditional bundle adjustment. Experiments on various datasets demonstrate the superiority of the proposed method in both camera tracking and mapping.

3D hand pose estimation is a challenging problem in computer vision due to the high degrees-of-freedom of hand articulated motion space and large viewpoint variation. As a consequence, similar poses observed from multiple views can be dramatically different. In order to deal with this issue, view-independent features are required to achieve state-of-the-art performance. In this paper, we investigate the impact of view-independent features on 3D hand pose estimation from a single depth image, and propose a novel recurrent neural network for 3D hand pose estimation, in which a cascaded 3D pose-guided alignment strategy is designed for view-independent feature extraction and a recurrent hand pose module is designed for modeling the dependencies among sequential aligned features for 3D hand pose estimation. In particular, our cascaded pose-guided 3D alignments are performed in 3D space in a coarse-to-fine fashion. First, hand joints are predicted and globally transformed into a canonical reference frame; Second, the palm of the hand is detected and aligned; Third, local transformations are applied to the fingers to refine the final predictions. The proposed recurrent hand pose module for aligned 3D representation can extract recurrent pose-aware features and iteratively refines the estimated hand pose. Our recurrent module could be utilized for both single-view estimation and sequence-based estimation with 3D hand pose tracking. Experiments show that our method improves the state-of-the-art by a large margin on popular benchmarks with the simple yet efficient alignment and network architectures.

Undesirable reflections contained in photos taken in front of glass windows or doors often degrade visual quality of the image. Separating two layers apart benefits both human and machine perception. The polarization status of the light changes after refraction or reflection, providing more observations of the scene, which can benefit the reflection separation. Different from previous works that take three or more polarization images as input, we propose to exploit physical constraints from a pair of unpolarized and polarized images to separate reflection and transmission layers in this paper. Due to the simplified capturing setup, the system is more under-determined compared to the existing polarization-based works. In order to solve this problem, we propose to estimate the semi-reflector orientation first to make the physical image formation well-posed, and then learn to reliably separate two layers using additional networks based on both physical and numerical analysis. In addition, a motion estimation network is introduced to handle the misalignment of paired input. Quantitative and qualitative experimental results show our approach performs favorably over existing polarization and single image based solutions.

We present a multi-sensor system for consistent 3D hand pose tracking and modeling that leverages the advantages of both wearable and optical sensors. Specifically, we employ a stretch-sensing soft glove and three IMUs in combination with an RGB-D camera. Different sensor modalities are fused based on the availability and confidence estimation, enabling seamless hand tracking in challenging environments with partial or even complete occlusion. To maximize the accuracy while maintaining high ease-of-use, we propose an automated user calibration that uses the RGB-D camera data to refine both the glove mapping model and the multi-IMU system parameters. Extensive experiments show that our setup outperforms the wearable-only approaches when the hand is in the field-of-view and outplays the camera-only methods when the hand is occluded.

This paper proposes a deep recurrent Rotation Averaging Graph Optimizer (RAGO) for Multiple Rotation Averaging (MRA). Conventional optimization-based methods usually fail to produce accurate results due to corrupted and noisy relative measurements. Recent learning-based approaches regard MRA as a regression problem, while these methods are sensitive to initialization due to the gauge freedom problem. To handle these problems, we propose a learnable iterative graph optimizer minimizing a gauge- invariant cost function with an edge rectification strategy to mitigate the effect of inaccurate measurements. Our graph optimizer iteratively refines the global camera rotations by minimizing each node's single rotation objective function. Besides, our approach iteratively rectifies relative rotations to make them more consistent with the current camera orientations and observed relative rotations. Furthermore,$we$employ a gated recurrent unit to improve the result by tracing the temporal information of the cost graph. Our framework is a real-time learning-to-optimize rotation averaging graph optimizer with a tiny size deployed for real-world applications. RAGO outperforms previous traditional and deep methods on real-world and synthetic datasets. The code is available at github.com/sfu-gruvi-3dv/RAGO.

Standard visual localization methods build a priori 3D model of a scene which is used to establish correspondences against the 2D keypoints in a query image. Storing these pre-built 3D scene models can be prohibitively expensive for large-scale environments, especially on mobile devices with limited storage and communication bandwidth. We design a novel framework that compresses a scene while still maintaining localization accuracy. The scene is compressed in three stages: first, the database frames are clustered using pairwise co-visibility information. Then, a learned point selection module prunes the points in each cluster taking into account the final pose estimation accuracy. In the final stage, the features of the selected points are further compressed using learned quantization. Query image registration is done using only the compressed scene points. To the best of our knowledge, we are the first to propose learned scene compression for visual localization. We also demonstrate the effectiveness and efficiency of our method on various outdoor datasets where it can perform accurate localization with low memory consumption.

Neural implicit representations have recently shown encouraging results in various domains, including promising progress in simultaneous localization and mapping (SLAM). Nevertheless, existing methods produce over- smoothed scene reconstructions and have difficulty scaling up to large scenes. These limitations are mainly due to their simple fully-connected network architecture that does not incorporate local information in the observations. In this paper, we present NICE-SLAM, a dense SLAM system that incorporates multi-level local information by introducing a hierarchical scene representation. Optimizing this representation with pre-trained geometric priors enables detailed reconstruction on large indoor scenes. Compared to recent neural implicit SLAM systems, our approach is more scalable, efficient, and robust. Experiments on five challenging datasets demonstrate competitive results of NICE-SLAM in both mapping and tracking quality. Project page: https://pengsongyou.github.io/nice-slam.

It is well known that visual SLAM systems based on dense matching are locally accurate but are also susceptible to long-term drift and map corruption. In contrast, feature matching methods can achieve greater long-term consistency but can suffer from inaccurate local pose estimation when feature information is sparse. Based on these observations, we propose an RGB-D SLAM system that leverages the advantages of both approaches: using dense frame-to-model odometry to build accurate sub-maps and on-the-fly feature-based matching across sub-maps for global map optimisation. In addition, we incorporate a learning-based loop closure component based on 3-D features which further stabilises map building. We have evaluated the approach on indoor sequences from public datasets, and the results show that it performs on par or better than state-of-the-art systems in terms of map reconstruction quality and pose estimation. The approach can also scale to large scenes where other systems often fail.

3D environment mapping has been actively stud-ied recently with the development of autonomous driving and augmented reality. Although many image-based methods are proposed due to their convenience and flexibility compared to other complex sensors, few works focus on fixing the inherent scale ambiguity of image-based methods and registering the reconstructed structure to the real-world 3D map, which is very important for autonomous driving. This paper presents a low-cost mapping solution that is able to refine and align the monocular reconstructed point cloud given a public street map. Specifically, we first find the association between the street map and the reconstructed point cloud structure by a novel graph-based geolocalization method. Then, optimized with the corresponding relationship, the map accuracy is significantly improved. The rich environment information can also be associated with the point cloud by the geographical location. Experiments show that our geolocalization algorithm can locate the scene on a gigantic city-scale map (173.46 km2) in two minutes and support 3D map reconstruction with absolute scale and rich environmental information from Internet videos.

Expanding an existing tourist photo from a partially captured scene to a full scene is one of the desired experiences for photography applications. Although photo extrapolation has been well studied, it is much more challenging to extrapolate a photo (i.e., selfie) from a narrow field of view to a wider one while maintaining a similar visual style. In this paper, we propose a factorized neural re-rendering model to produce photorealistic novel views from cluttered outdoor Internet photo collections, which enables the applications including controllable scene re-rendering, photo extrapolation and even extrapolated 3D photo generation. Specifically, we first develop a novel factorized re-rendering pipeline to handle the ambiguity in the decomposition of geometry, appearance and illumination. We also propose a composited training strategy to tackle the unexpected occlusion in Internet images. Moreover, to enhance photo-realism when extrapolating tourist photographs, we propose a novel realism augmentation process to complement appearance details, which automatically propagates the texture details from a narrow captured photo to the extrapolated neural rendered image. The experiments and photo editing examples on outdoor scenes demonstrate the superior performance of our proposed method in both photo-realism and downstream applications. Code and the supplementary material are available on the project webpage: https://zju3dv.github.io/neural_outdoor_rerender/.

We, as human beings, can understand and picture a familiar scene from arbitrary viewpoints given a single image, whereas this is still a grand challenge for computers. We hereby present a novel solution to mimic such human perception capability based on a new paradigm of amodal 3D scene understanding with neural rendering for a closed scene. Specifically, we first learn the prior knowledge of the objects in a closed scene via an offline stage, which facilitates an online stage to understand the room with unseen furniture arrangement. During the online stage, given a panoramic image of the scene in different layouts, we utilize a holistic neural-rendering-based optimization framework to efficiently estimate the correct 3D scene layout and deliver realistic free-viewpoint rendering. In order to handle the domain gap between the offline and online stage, our method exploits compositional neural rendering techniques for data augmentation in the offline training. The experiments on both synthetic and real datasets demonstrate that our two-stage design achieves robust 3D scene understanding and outperforms competing methods by a large margin, and we also show that our realistic free-viewpoint rendering enables various applications, including scene touring and editing. Code and data are available on the project webpage: https://zju3dv.github.io/nr_in_a_room/.

Bundle adjustment (BA) is widely used in SLAM and SfM, which are key technologies in Augmented Reality. For real-time SLAM and large-scale SfM, the efficiency of BA is of great importance. This paper proposes CoLi-BA, a novel and efficient BA solver that significantly improves the optimization speed by compact linearization and reordering. Specifically, for each reprojection function, the redundant matrix representation of Jacobian is replaced with a tiny 3D vector, by which the computational complexity, memory storage, and cache missing for Hessian matrix construction and Schur complement are significantly reduced. Besides, we also propose a novel reordering strategy to improve the cache efficiency for Schur complement. Experiments on diverse datasets show that the speed of the proposed CoLi-BA is five times that of Ceres and two times that of g2o without sacrificing accuracy. We further verify the effectiveness by porting CoLi-BA to the open-source SLAM and SfM systems. Even when running the proposed solver in a single thread, the local BA of SLAM only takes about 20ms on a desktop PC, and the reconstruction of SfM with seven thousand photos only takes half an hour. The source code is available on the webpage: https://github.com/zju3dv/CoLi-BA.

We introduce a novel task of reconstructing a time series of second-person13D human body meshes from monocular egocentric videos. The unique viewpoint and rapid embodied camera motion of egocentric videos raise additional technical barriers for human body capture. To address those challenges, we propose a simple yet effective optimization-based approach that leverages 2D observations of the entire video sequence and human-scene interaction constraint to estimate second-person human poses, shapes, and global motion that are grounded on the 3D environment captured from the egocentric view. We conduct detailed ablation studies to validate our design choice. Moreover, we compare our method with the previous state-of-the-art method on human motion capture from monocular video, and show that our method estimates more accurate human-body poses and shapes under the challenging egocentric setting. In addition, we demonstrate that our approach produces more realistic human-scene interaction.

3D representation and reconstruction of human bodies have been studied for a long time in computer vision. Traditional methods rely mostly on parametric statistical linear models, limiting the space of possible bodies to linear combinations. It is only recently that some approaches try to leverage neural implicit representations for human body modeling, and while demonstrating impressive results, they are either limited by representation capability or not physically meaningful and controllable. In this work, we propose a novel neural implicit representation for the human body, which is fully differentiable and optimizable with disentangled shape and pose latent spaces. Contrary to prior work, our representation is designed based on the kinematic model, which makes the representation controllable for tasks like pose animation, while simultaneously allowing the optimization of shape and pose for tasks like 3D fitting and pose tracking. Our model can be trained and fine-tuned directly on non-watertight raw data with well-designed losses. Experiments demonstrate the improved 3D reconstruction performance over SoTA approaches and show the applicability of our method to shape interpolation, model fitting, pose tracking, and motion retargeting.

This paper presents a method for riggable 3D face reconstruction from monocular images, which jointly estimates a personalized face rig and per-image parameters including expressions, poses, and illuminations. To achieve this goal, we design an end-to-end trainable network embedded with a differentiable in-network optimization. The network first parameterizes the face rig as a compact latent code with a neural decoder, and then estimates the latent code as well as per-image parameters via a learnable optimization. By estimating a personalized face rig, our method goes beyond static reconstructions and enables downstream applications such as video retargeting. In-network optimization explicitly enforces constraints derived from the first principles, thus introduces additional priors than regression-based methods. Finally, data-driven priors from deep learning are utilized to constrain the ill-posed monocular setting and ease the optimization difficulty. Experiments demonstrate that our method achieves SOTA reconstruction accuracy, reasonable robustness and generalization ability, and supports standard face rig applications.

This paper presents an end-to-end neural network for multiple rotation averaging in SfM. Due to the manifold constraint of rotations, conventional methods usually take two separate steps involving spanning tree based initialization and iterative nonlinear optimization respectively. These methods can suffer from bad initializations due to the noisy spanning tree or outliers in input relative rotations. To handle these problems, we propose to integrate initialization and optimization together in an unified graph neural network via a novel differentiable multi-source propagation module. Specifically, our network utilizes the image context and geometric cues in feature correspondences to reduce the impact of outliers. Furthermore, unlike the methods that utilize the spanning tree to initialize orientations according to a single reference node in a top-down manner, our net-work initializes orientations according to multiple sources while utilizing information from all neighbors in a differentiable way. More importantly, our end-to-end formulation also enables iterative re-weighting of input relative orientations at test time to improve the accuracy of the final estimation by minimizing the impact of outliers. We demonstrate the effectiveness of our method on two real-world datasets, achieving state-of-the-art performance.

We present a new pipeline for holistic 3D scene understanding from a single image, which could predict object shapes, object poses, and scene layout. As it is a highly ill-posed problem, existing methods usually suffer from inaccurate estimation of both shapes and layout especially for the cluttered scene due to the heavy occlusion between objects. We propose to utilize the latest deep implicit representation to solve this challenge. We not only propose an image-based local structured implicit network to improve the object shape estimation, but also refine the 3D object pose and scene layout via a novel implicit scene graph neural network that exploits the implicit local object features. A novel physical violation loss is also proposed to avoid incorrect context between objects. Extensive experiments demonstrate that our method outperforms the state-of-the-art methods in terms of object shape, scene layout estimation, and 3D object detection.

Accurately describing and detecting 2D and 3D key-points is crucial to establishing correspondences across images and point clouds. Despite a plethora of learning-based 2D or 3D local feature descriptors and detectors having been proposed, the derivation of a shared descriptor and joint keypoint detector that directly matches pixels and points remains under-explored by the community. This work takes the initiative to establish fine-grained correspondences between 2D images and 3D point clouds. In order to directly match pixels and points, a dual fully-convolutional framework is presented that maps 2D and 3D inputs into a shared latent representation space to simultaneously describe and detect keypoints. Furthermore, an ultra-wide reception mechanism and a novel loss function are designed to mitigate the intrinsic information variations between pixel and point local regions. Extensive experimental results demonstrate that our framework shows competitive performance in fine-grained matching between images and point clouds and achieves state-of-the-art results for the task of indoor visual localization. Our source code is available at https://github.com/BingCS/P2-Net.

We aim at improving the computational efficiency of graph convolutional networks (GCNs) for learning on point clouds. The basic graph convolution that is composed of a K-nearest neighbor (KNN) search and a multilayer perceptron (MLP) is examined. By mathematically analyzing the operations there, two findings to improve the efficiency of GCNs are obtained. (1) The local geometric structure information of 3D representations propagates smoothly across the GCN that relies on KNN search to gather neighborhood features. This motivates the simplification of multiple KNN searches in GCNs. (2) Shuffling the order of graph feature gathering and an MLP leads to equivalent or similar composite operations. Based on those findings, we optimize the computational procedure in GCNs. A series of experiments show that the optimized networks have reduced computational complexity, decreased memory consumption, and accelerated inference speed while maintaining comparable accuracy for learning on point clouds.

We present a novel method for synthesizing both temporally and geometrically consistent street-view panoramic video from a single satellite image and camera trajectory. Existing cross-view synthesis approaches focus on images, while video synthesis in such a case has not yet received enough attention. For geometrical and temporal consistency, our approach explicitly creates a 3D point cloud representation of the scene and maintains dense 3D-2D correspondences across frames that reflect the geometric scene configuration inferred from the satellite view. As for synthesis in the 3D space, we implement a cascaded network architecture with two hourglass modules to generate point-wise coarse and fine features from semantics and per-class latent vectors, followed by projection to frames and an up-sampling module to obtain the final realistic video. By leveraging computed correspondences, the produced street-view video frames adhere to the 3D geometric scene structure and maintain temporal consistency. Qualitative and quantitative experiments demonstrate superior results compared to other state-of-the-art synthesis approaches that either lack temporal consistency or realistic appearance. To the best of our knowledge, our work is the first one to synthesize cross-view images to videos..

Different from traditional video cameras, event cam- eras capture asynchronous events stream in which each event encodes pixel location, trigger time, and the polarity of the brightness changes. In this paper, we introduce a novel graph-based framework for event cameras, namely SlideGCN. Unlike some recent graph-based methods that use groups of events as input, our approach can efficiently process data event-by-event, unlock the low latency nature of events data while still maintaining the graph’s structure internally. For fast graph construction, we develop a radius search algorithm, which better exploits the partial regular structure of event cloud against k-d tree based generic methods. Experiments show that our method reduces the computational complexity up to 100 times with respect to current graph-based methods while keeping state-of-the-art performance on object recognition. Moreover, we verify the superiority of event-wise processing with our method. When the state becomes stable, we can give a prediction with high confidence, thus making an early recognition.

We present a novel framework for mesh reconstruction from unstructured point clouds by taking advantage of the learned visibility of the 3D points in the virtual views and traditional graph-cut based mesh generation. Specifically, we first propose a three-step network that explicitly employs depth completion for visibility prediction. Then the visibility information of multiple views is aggregated to generate a 3D mesh model by solving an optimization problem considering visibility in which a novel adaptive visibility weighting in surface determination is also introduced to suppress line of sight with a large incident angle. Compared to other learning-based approaches, our pipeline only exercises the learning on a 2D binary classification task, i.e., points visible or not in a view, which is much more generalizable and practically more efficient and capable to deal with a large number of points. Experiments demonstrate that our method with favorable transferability and robustness, and achieve competing performances w.r.t. state-of-the-art learning-based approaches on small complex objects and outperforms on large indoor and outdoor scenes. Code is available at https://github.com/GDAOSU/vis2mesh.

We present a deep learning pipeline that leverages network self-prior to recover a full 3D model consisting of both a triangular mesh and a texture map from the colored 3D point cloud. Different from previous methods either exploiting 2D self-prior for image editing or 3D self-prior for pure surface reconstruction, we propose to exploit a novel hybrid 2D-3D self-prior in deep neural networks to significantly improve the geometry quality and produce a high-resolution texture map, which is typically missing from the output of commodity-level 3D scanners. In particular, we first generate an initial mesh using a 3D convolutional neural network with 3D self-prior, and then encode both 3D information and color information in the 2D UV atlas, which is further refined by 2D convolutional neural networks with the self-prior. In this way, both 2D and 3D self-priors are utilized for the mesh and texture recovery. Experiments show that, without the need of any additional training data, our method recovers the 3D textured mesh model of high quality from sparse input, and outperforms the state-of-the-art methods in terms of both geometry and texture quality.

Implicit neural rendering techniques have shown promising results for novel view synthesis. However, existing methods usually encode the entire scene as a whole, which is generally not aware of the object identity and limits the ability to the high-level editing tasks such as moving or adding furniture. In this paper, we present a novel neural scene rendering system, which learns an object-compositional neural radiance field and produces realistic rendering with editing capability for a clustered and real-world scene. Specifically, we design a novel two-pathway architecture, in which the scene branch encodes the scene geometry and appearance, and the object branch encodes each standalone object conditioned on learnable object activation codes. To survive the training in heavily cluttered scenes, we propose a scene-guided training strategy to solve the 3D space ambiguity in the occluded regions and learn sharp boundaries for each object. Extensive experiments demonstrate that our system not only achieves competitive performance for static scene novel-view synthesis, but also produces realistic rendering for object-level editing.

Panorama images have a much larger field-of-view thus naturally encode enriched scene context information compared to standard perspective images, which however is not well exploited in the previous scene understanding methods. In this paper, we propose a novel method for panoramic 3D scene understanding which recovers the 3D room layout and the shape, pose, position, and semantic category for each object from a single full-view panorama image. In order to fully utilize the rich context information, we design a novel graph neural network based context model to predict the relationship among objects and room layout, and a differentiable relationship-based optimization module to optimize object arrangement with well-designed objective functions on-the-fly. Realizing the existing data are either with incomplete ground truth or overly-simplified scene, we present a new synthetic dataset with good diversity in room layout and furniture placement, and realistic image quality for total panoramic 3D scene understanding. Experiments demonstrate that our method outperforms existing methods on panoramic scene understanding in terms of both geometry accuracy and object arrangement. Code is available at https://chengzhag.github.io/publication/dpc.

Real-time dense reconstruction has been extensively studied for its wide applications in computer vision and robotics, meanwhile much effort has been made for the multi-robot system which plays an irreplaceable role in complicated but time-critical scenarios, e.g., search and rescue tasks. In this paper, we propose an efficient system named Coxgraph for multi-robot collaborative dense reconstruction in real-time. In our system, each client performs volumetric mapping in a producer-consumer manner. To facilitate transmission, we propose a compact 3D representation which transforms the SDF submap to mesh packs. During the recovery of submaps from mesh packs, the system can perform loop closure outlier rejection based on geometry consistency, trajectory collision and fitness check. Then we develop a robust map fusion method through joint optimization of trajectories and submaps. Extensive experiments demonstrate that our system can produce a globally consistent dense map in real-time with less transmission load, which is available as open-source software1.

Robust and accurate global 6DoF localization is essential for many applications, i.e., augmented reality and autonomous driving. Most existing 6DoF visual localization approaches need to build a dense texture model in advance, which is computationally extensive and almost infeasible in the global range. In this work, we propose BDLoc, a hierarchical global localization framework via the 2.5D building map, which is able to estimate the accurate pose of the query street-view image without using detailed dense 3D model and texture information. Specifically speaking, we first extract the 3D building information from the street-view image and surrounding 2.5D building map, and then solve a coarse relative pose by local to global registration. In order to improve the feature extraction, we propose a novel SPG-Net which is able to capture both local and global features. Finally, an iterative semantic alignment is applied to obtain a finner result with the differentiable rendering and the cross-view semantic constraint. Except for a coarse longitude and latitude from GPS, BDLoc doesn’t need any additional information like altitude and orientation that are necessary for many previous works. We also create a large dataset to explore the performance of the 2.5D map-based localization task. Extensive experiments demonstrate the superior performance of our method.

Hand pose understanding is essential to applications such as human computer interaction and augmented reality. Recently, deep learning based methods achieve great progress in this problem. However, the lack of high-quality and large-scale dataset prevents the further improvement of hand pose related tasks such as 2D/3D hand pose from color and depth from color. In this paper, we develop a large-scale and high-quality synthetic dataset, PBRHand. The dataset contains millions of photo-realistic rendered hand images and various ground truths including pose, semantic segmentation, and depth. Based on the dataset, we firstly investigate the effect of rendering methods and used databases on the performance of three hand pose related tasks: 2D/3D hand pose from color, depth from color and 3D hand pose from depth. This study provides insights that photo-realistic rendering dataset is worthy of synthesizing and shows that our new dataset can improve the performance of the state-of-the-art on these tasks. This synthetic data also enables us to explore multi-task learning, while it is expensive to have all the ground truth available on real data. Evaluations show that our approach can achieve state-of-the-art or competitive performance on several public datasets.

Recent emerging technologies such AR/VR and HCI are drawing high demand on more comprehensive hand shape understanding, requiring not only 3D hand skeleton pose but also hand shape geometry. In this paper, we propose a deep learning framework to produce 3D hand shape from a single depth image. To address the challenge that capturing ground truth 3D hand shape in the training dataset is non-trivial, we leverage synthetic data to construct a statistical hand shape model and adopt weak supervision from widely accessible hand skeleton pose annotation. To bridge the gap due to the different hand skeleton definitions in the existing public datasets, we propose a joint regression network for hand pose adaptation. To reconstruct the hand shape, we use Chamfer loss between the predicted hand shape and the point cloud from the input depth to learn the shape reconstruction model in a weakly-supervised manner. Experiments demonstrate that our model adapts well to the real data and produces accurate hand shapes that outperform the state-of-the-art methods both qualitatively and quantitatively.

We present a method for 3D face reconstruction from multi-view images with different expressions. We formulate this problem from the perspective of non-rigid multi-view stereo (NRMVS). Unlike previous learning-based methods, which often regress the face shape directly, our method optimizes the 3D face shape by explicitly enforcing multi-view appearance consistency, which is known to be effective in recovering shape details according to conventional multi-view stereo methods. Furthermore, by estimating face shape through optimization based on multi-view consistency, our method can potentially have better generalization to unseen data. However, this optimization is challenging since each input image has a different expression. We facilitate it with a CNN network that learns to regularize the non-rigid 3D face according to the input image and preliminary optimization results. Extensive experiments show that our method achieves the state-of-the-art performance on various datasets and generalizes well to in-the-wild data.

We present a novel network for rolling shutter effect correction. Our network takes two consecutive rolling shutter images and estimates the corresponding global shutter image of the latest frame. The dense displacement field from a rolling shutter image to its corresponding global shutter image is estimated via a motion estimation network. The learned feature representation of a rolling shutter image is then warped, via the displacement field, to its global shutter representation by a differentiable forward warping block. An image decoder recovers the global shutter image based on the warped feature representation. Our network can be trained end-to-end and only requires the global shutter image for supervision. Since there is no public dataset available, we also propose two large datasets: the Carla-RS dataset and the Fastec-RS dataset. Experimental results demonstrate that our network outperforms the state-of-the-art methods. We make both our code and datasets available at https://github.com/ethliup/DeepUnrollNet.

We propose a differentiable sphere tracing algorithm to bridge the gap between inverse graphics methods and the recently proposed deep learning based implicit signed distance function. Due to the nature of the implicit function, the rendering process requires tremendous function queries, which is particularly problematic when the function is represented as a neural network. We optimize both the forward and backward pass of our rendering layer to make it run efficiently with affordable memory consumption on a commodity graphics card. Our rendering method is fully differentiable such that losses can be directly computed on the rendered 2D observations, and the gradients can be propagated backward to optimize the 3D geometry. We show that our rendering method can effectively reconstruct accurate 3D shapes from various inputs, such as sparse depth and multi-view images, through inverse optimization. With the geometry based reasoning, our 3D shape prediction methods show excellent generalization capability and robustness against various noises.

We present a novel method for generating panoramic street-view images which are geometrically consistent with a given satellite image. Different from existing approaches that completely rely on a deep learning architecture to generalize cross-view image distributions, our approach explicitly loops in the geometric configuration of the ground objects based on the satellite views, such that the produced ground view synthesis preserves the geometric shape and the semantics of the scene. In particular, we propose a neural network with a geo-transformation layer that turns predicted ground-height values from the satellite view to a ground view while retaining the physical satellite-to-ground relation. Our results show that the synthesized image retains well-articulated and authentic geometric shapes, as well as texture richness of the street-view in various scenarios. Both qualitative and quantitative results demonstrate that our method compares favorably to other state-of-the-art approaches that lack geometric consistency.

Previous methods on estimating detailed human depth often require supervised training with ‘ground truth’ depth data. This paper presents a self-supervised method that can be trained on YouTube videos without known depth, which makes training data collection simple and improves the generalization of the learned network. The self-supervised learning is achieved by minimizing a photo-consistency loss, which is evaluated between a video frame and its neighboring frames warped according to the estimated depth and the 3D non-rigid motion of the human body. To solve this non-rigid motion, we first estimate a rough SMPL model at each video frame and compute the non-rigid body motion accordingly, which enables self-supervised learning on estimating the shape details. Experiments demonstrate that our method enjoys better generalization, and performs much better on data in the wild.

In this paper, we present an omnidirectional localization and dense mapping system for a wide-baseline multiview stereo setup with ultra-wide field-of-view (FOV) fisheye cameras, which has a 360° coverage of stereo observations of the environment. For more practical and accurate reconstruction, we first introduce improved and light-weighted deep neural networks for the omnidirectional depth estimation, which are faster and more accurate than the existing networks. Second, we integrate our omnidirectional depth estimates into the visual odometry (VO) and add a loop closing module for global consistency. Using the estimated depth map, we reproject keypoints onto each other view, which leads to a better and more efficient feature matching process. Finally, we fuse the omnidirectional depth maps and the estimated rig poses into the truncated signed distance function (TSDF) volume to acquire a 3D map. We evaluate our method on synthetic datasets with ground-truth and real-world sequences of challenging environments, and the extensive experiments show that the proposed system generates excellent reconstruction results in both synthetic and real-world environments.

In this paper, we propose a deep learning architecture that produces accurate dense depth for the outdoor scene from a single color image and a sparse depth. Inspired by the indoor depth completion, our network estimates surface normals as the intermediate representation to produce dense depth, and can be trained end-to-end. With a modified encoder-decoder structure, our network effectively fuses the dense color image and the sparse LiDAR depth. To address outdoor specific challenges, our network predicts a confidence mask to handle mixed LiDAR signals near foreground boundaries due to occlusion, and combines estimates from the color image and surface normals with learned attention maps to improve the depth accuracy especially for distant areas. Extensive experiments demonstrate that our model improves upon the state-of-the-art performance on KITTI depth completion benchmark. Ablation study shows the positive impact of each model components to the final performance, and comprehensive analysis shows that our model generalizes well to the input with higher sparsity or from indoor scenes.

In this paper we consider the problem of relative pose estimation from two images with per-pixel polarimetric information. Using these additional measurements we derive a simple minimal solver for the essential matrix which only requires two point correspondences. The polarization constraints allow us to pointwise recover the 3D surface normal up to a two-fold ambiguity for the diffuse reflection. Since this ambiguity exists per point, there is a combinatorial explosion of possibilities. However, since our solver only requires two point correspondences, we only need to consider 16 configurations when solving for the relative pose. Once the relative orientation is recovered, we show that it is trivial to resolve the ambiguity for the remaining points. For robustness, we also propose a joint optimization between the relative pose and the refractive index to handle the refractive distortion. In experiments, on both synthetic and real data, we demonstrate that by leveraging the additional information available from polarization cameras, we can improve over classical methods which only rely on the 2D-point locations to estimate the geometry. Finally, we demonstrate the practical applicability of our approach by integrating it into a state-of-the-art global Structure-from-Motion pipeline.

We present a real-time dense geometric mapping algorithm for large-scale environments. Unlike existing methods which use pinhole cameras, our implementation is based on fisheye cameras whose large field of view benefits various computer vision applications for self-driving vehicles such as visual-inertial odometry, visual localization, and object detection. Our algorithm runs on in-vehicle PCs at approximately 15 Hz, enabling vision-only 3D scene perception for self-driving vehicles. For each synchronized set of images captured by multiple cameras, we first compute a depth map for a reference camera using plane-sweeping stereo. To maintain both accuracy and efficiency, while accounting for the fact that fisheye images have a lower angular resolution, we recover the depths using multiple image resolutions. We adopt the fast object detection framework, YOLOv3, to remove potentially dynamic objects. At the end of the pipeline, we fuse the fisheye depth images into the truncated signed distance function (TSDF) volume to obtain a 3D map. We evaluate our method on large-scale urban datasets, and results show that our method works well in complex dynamic environments.

Visual localization, i.e., determining the position and orientation of a vehicle with respect to a map, is a key problem in autonomous driving. We present a multi-camera visual inertial localization algorithm for large scale environments. To efficiently and effectively match features against a pre-built global 3D map, we propose a prioritized feature matching scheme for multi-camera systems. In contrast to existing works, designed for monocular cameras, we (1) tailor the prioritization function to the multi-camera setup and (2) run feature matching and pose estimation in parallel. This significantly accelerates the matching and pose estimation stages and allows us to dynamically adapt the matching efforts based on the surrounding environment. In addition, we show how pose priors can be integrated into the localization system to increase efficiency and robustness. Finally, we extend our algorithm by fusing the absolute pose estimates with motion estimates from a multi-camera visual inertial odometry pipeline (VIO). This results in a system that provides reliable and drift-less pose estimation. Extensive experiments show that our localization runs fast and robust under varying conditions, and that our extended algorithm enables reliable real-time pose estimation.

Project AutoVision aims to develop localization and 3D scene perception capabilities for a self-driving vehicle. Such capabilities will enable autonomous navigation in urban and rural environments, in day and night, and with cameras as the only exteroceptive sensors. The sensor suite employs many cameras for both 360-degree coverage and accurate multi-view stereo; the use of low-cost cameras keeps the cost of this sensor suite to a minimum. In addition, the project seeks to extend the operating envelope to include GNSS-less conditions which are typical for environments with tall buildings, foliage, and tunnels. Emphasis is placed on leveraging multi-view geometry and deep learning to enable the vehicle to localize and perceive in 3D space. This paper presents an overview of the project, and describes the sensor suite and current progress in the areas of calibration, localization, and perception.

When we take photos through glass windows or doors, the transmitted background scene is often blended with undesirable reflection. Separating two layers apart to enhance the image quality is of vital importance for both human and machine perception. In this paper, we propose to exploit physical constraints from a pair of unpolarized and polarized images to separate reflection and transmission layers. Due to the simplified capturing setup, the system becomes more underdetermined compared with existing polarization based solutions that take three or more images as input. We propose to solve semireflector orientation estimation first to make the physical image formation well-posed and then learn to reliably separate two layers using a refinement network with gradient loss. Quantitative and qualitative experimental results show our approach performs favorably over existing polarization and single image based solutions.

This paper presents a novel polarimetric dense monocular SLAM (PDMS) algorithm based on a polarization camera. The algorithm exploits both photometric and polarimetric light information to produce more accurate and complete geometry. The polarimetric information allows us to recover the azimuth angle of surface normals from each video frame to facilitate dense reconstruction, especially at textureless or specular regions. There are two challenges in our approach: 1) surface azimuth angles from the polarization camera are very noisy; and 2) we need a near real-time solution for SLAM. Previous successful methods on polarimetric multi-view stereo are offline and require manually pre-segmented object masks to suppress the effects of erroneous angle information along boundaries. Our fully automatic approach efficiently iterates azimuth-based depth propagations, two-view depth consistency check, and depth optimization to produce a depthmap in real-time, where all the algorithmic steps are carefully designed to enable a GPU implementation. To our knowledge, this paper is the first to propose a photometric and polarimetric method for dense SLAM. We have qualitatively and quantitatively evaluated our algorithm against a few of competing methods, demonstrating the superior performance on various indoor and outdoor scenes.

Multi-view stereo relies on feature correspondences for 3D reconstruction, and thus is fundamentally flawed in dealing with featureless scenes. In this paper, we propose polarimetric multi-view stereo, which combines per-pixel photometric information from polarization with epipolar constraints from multiple views for 3D reconstruction. Polarization reveals surface normal information, and is thus helpful to propagate depth to featureless regions. Polarimetric multi-view stereo is completely passive and can be applied outdoors in uncontrolled illumination, since the data capture can be done simply with either a polarizer or a polarization camera. Unlike previous work on shape-from-polarization which is limited to either diffuse polarization or specular polarization only, we propose a novel polarization imaging model that can handle real-world objects with mixed polarization. We prove there are exactly two types of ambiguities on estimating surface azimuth angles from polarization, and we resolve them with graph optimization and iso-depth contour tracing. This step significantly improves the initial depth map estimate, which are later fused together for complete 3D reconstruction. Extensive experimental results demonstrate high-quality 3D reconstruction and better performance than state-of-the-art multi-view stereo methods, especially on featureless 3D objects, such as ceramic tiles, office room with white walls, and highly reflective cars in the outdoors.

Time slice photography is a popular effect that visualizes the passing of time by aligning and stitching multiple images capturing the same scene at different times together into a single image. Extending this effect to video is a difficult problem, and one where existing solutions have only had limited success. In this paper, we propose an easy-to-use and robust system for creating time slice videos from a wide variety of consumer videos. The main technical challenge we address is how to align videos taken at different times with substantially different appearances, in the presence of moving objects and moving cameras with slightly different trajectories. To achieve a temporally stable alignment, we perform a mixed 2D-3D alignment, where a rough 3D reconstruction is used to generate sparse constraints that are integrated into a pixelwise 2D registration. We apply our method to a number of challenging scenarios, and show that we can achieve a higher quality registration than prior work. We propose a 3D user interface that allows the user to easily specify how multiple videos should be composited in space and time. Finally, we show that our alignment method can be applied in more general video editing and compositing tasks, such as object removal.

Global structure-from-motion (SfM) algorithms register all cameras simultaneously, which are potentially more efficient and less prone to drifting than incremental SfM methods. Global SfM methods often solve the camera orientations and positions separately. This paper focuses on the problem of global position (i.e. translation) estimation. Essential matrix based global translation estimation methods (e.g. [1]) usually degenerate at collinear camera motion because the translation scale is not determined by an essential matrix. Trifocal tensor based methods (e.g. [3]) usually rely on a strongly connected camera-triplet graph, where two triplets are connected by their common edge. The 3D reconstruction will distort or break into disconnected components when such strong association among images does not exist. The recent 1DSfM method [4] designs a smart filter to discard outlier essential matrices and solves scene points and cameras together by enforcing orientation consistency. However, this method requires abundant association between input images, e.g.∼O(n2) essential matrices for n cameras, which is more suitable for Internet images and often fails on sequentially captured data. The data association problem of [4] and [3] is exemplified in Figure 1. The Street example on the top is a sequential data where each image is only matched upto 4 neighbors. 1DSfM fails on this example due to insufficient image association. In the Seville example on the bottom, those Internet images are mostly captured from two viewpoints (see the two representative sample images) with weak affinity between images at different viewpoints. This weak data association causes seriously distorted reconstruction for the triplet-based method in [3]. This paper introduces a direct linear algorithm to address the presented challenges. It avoids degeneracy at collinear motion and deals with weakly associated data. Our method capitalizes on constraints from essential matrices and feature tracks. As shown in Figure 2 (a), the location of a scene point p can be computed as the middle point of the mutual perpendicular line segment AB of the two rays passing through p’s image projections:

Global structure-from-motion (SfM) methods solve all cameras simultaneously from all available relative motions. It has better potential in both reconstruction accuracy and computation efficiency than incremental methods. However, global SfM is challenging, mainly because of two reasons. Firstly, translation averaging is difficult, since an essential matrix only tells the direction of relative translation. Secondly, it is also hard to filter out bad essential matrices due to feature matching failures. We propose to compute a sparse depth image at each camera to solve both problems. Depth images help to upgrade an essential matrix to a similarity transformation, which can determine the scale of relative translation. Thus, camera registration is formulated as a well-posed similarity averaging problem. Depth images also make the filtering of essential matrices simple and effective. In this way, translation averaging can be solved robustly in two convex L1 optimization problems, which reach the global optimum rapidly. We demonstrate this method in various examples including sequential data, Internet data, and ambiguous data with repetitive scene structures.

We present a linear method for global camera pose registration from pair wise relative poses encoded in essential matrices. Our method minimizes an approximate geometric error to enforce the triangular relationship in camera triplets. This formulation does not suffer from the typical `unbalanced scale' problem in linear methods relying on pair wise translation direction constraints, i.e. an algebraic error, nor the system degeneracy from collinear motion. In the case of three cameras, our method provides a good linear approximation of the trifocal tensor. It can be directly scaled up to register multiple cameras. The results obtained are accurate for point triangulation and can serve as a good initialization for final bundle adjustment. We evaluate the algorithm performance with different types of data and demonstrate its effectiveness. Our system produces good accuracy, robustness, and outperforms some well-known systems on efficiency.

There lies a geometric relationship (homography) between two frames in the video sequences captured by pan-tilt (PT) cameras due to their constrained movement, and the geometric relationship is valuable for reducing the spatial redundancy in video coding. In this paper, we propose a novel homography-based search (HBS) algorithm for block motion estimation in coding the sequences captured by PT cameras, which well utilizes the homography between two frames. In addition, adaptive thresholds are adopted in our method to classify different kinds of blocks. Compared with other traditional fast algorithms, the proposed HBS algorithm is proved more efficient for the sequences captured by PT cameras.

An adaptive threshold for early termination is proposed in this paper to enhance the cross octagonal search algorithm (COSA). In block motion estimation, search patterns have a very important impact on searching speed and distortion performance. COSA achieves almost the same visual quality with full-search algorithm while it takes much fewer search points than hybrid unsymmetrical-cross multi-hexagon-grid search (UMHexagonS) indeed. An adaptive threshold for early termination is introduced to COSA which avoids meaningless calculation after the searching point is good enough. The simulation results show that the proposed method reduces the motion estimation time by from 10.90% to 42.75% than UMHexagonS according to different types of sequences with negligible coding loss.