Class-agnostic 3D instance segmentation tackles the challenging task of segmenting all object instances, including previously unseen ones, without semantic class reliance. Current methods struggle with generalization due to the scarce annotated 3D scene data or noisy 2D segmentations. While synthetic data generation offers a promising solution, existing 3D scene synthesis methods fail to simultaneously satisfy geometry diversity, context complexity, and layout reasonability, each essential for this task. To address these needs, we propose an Adapted 3D Scene Synthesis pipeline for class-agnostic 3D Instance SegmenTation, termed as ASSIST-3D, to synthesize proper data for model generalization enhancement. Specifically, ASSIST-3D features three key innovations, including 1) Heterogeneous Object Selection from extensive 3D CAD asset collections, incorporating randomness in object sampling to maximize geometric and contextual diversity; 2) Scene Layout Generation through LLM-guided spatial reasoning combined with depth-first search for reasonable object placements; and 3) Realistic Point Cloud Construction via multi-view RGB-D image rendering and fusion from the synthetic scenes, closely mimicking real-world sensor data acquisition. Experiments on ScanNetV2, ScanNet++, and S3DIS benchmarks demonstrate that models trained with ASSIST-3D-generated data significantly outperform existing methods. Further comparisons underscore the superiority of our purpose-built pipeline over existing 3D scene synthesis approaches.

Sitong Wu, Haoru Tan, Xichen Zhang, Bin Xia, Wenhu Zhang, Xiaojuan Qi, Bei Yu, Jiaya Jia. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Human avatars, defined as animatable 3D digital people, have diverse applications, including teleconferencing, motion guidance, remote care, and gaming. With the widespread adoption of smart mobile devices, the demand for real-time viewing of dynamic digital humans on mobile platforms has grown significantly. However, existing avatar reconstruction methods, such as MagicStream and ExAvatar, struggle to produce high-fidelity avatars from monocular videos captured by mobile devices. Furthermore, most related research focuses on individual algorithmic components and does not provide an end-to-end pipeline that jointly addresses avatar reconstruction, 4D data transmission, and mobile rendering. In this work, we introduce HoloMobile, the first end-to-end mobile avatar system with two key capabilities: (1) High-fidelity avatar reconstruction from monocular videos captured on mobile phones, using a hybrid explicit-implicit model. (2) Efficient polynomial compression, transmission, and web-based rendering of dynamic 3D avatars on mobile devices. With avatar reconstruction performed on a PC or cloud with a single consumer-grade GPU, HoloMobile provides an end-to-end solution for mobile users: from data capture to streaming and real-time viewing. Evaluation on 20 self-captured avatars reveals that HoloMobile achieves an average PSNR of 30.36 dB, a compression rate of 99.57%, and over 200 FPS for mobile rendering, demonstrating superiority over existing baselines. The demo video of HoloMobile is available at https://youtu.be/kfOFV7AiP5A.

Mixture-of-Experts (MoE) has emerged as an effective and efficient scaling mechanism for large language models (LLMs) and vision-language models (VLMs). By expanding a single feed-forward network into multiple expert branches, MoE increases model capacity while maintaining efficiency through sparse activation. However, despite this sparsity, the need to preload all experts into memory and activate multiple experts per input introduces significant computational and memory overhead. The expert module becomes the dominant contributor to model size and inference cost, posing a major challenge for deployment. To address this, we propose MC# (Mixture-Compressor-sharp), a unified framework that combines static quantization and dynamic expert pruning by leveraging the significance of both experts and tokens to achieve aggressive compression of MoE-LLMs/VLMs. To reduce storage and loading overhead, we introduce Pre-Loading Mixed-Precision Quantization (PMQ), which formulates adaptive bit allocation as a linear programming problem. The objective function jointly considers expert importance and quantization error, producing a Pareto-optimal trade-off between model size and performance. To reduce runtime computation, we further introduce Online Top-any Pruning (OTP), which models expert activation per token as a learnable distribution via Gumbel-Softmax sampling. During inference, OTP dynamically selects a subset of experts for each token, allowing fine-grained control over activation. By combining PMQ's static bit-width optimization with OTP's dynamic routing, MC# achieves extreme compression with minimal accuracy degradation. On DeepSeek-VL2, MC# achieves a 6.2 × weight reduction at an average of 2.57 bits, with only a 1.7% drop across five multimodal benchmarks compared to the 16-bit baseline. Moreover, OTP further reduces expert activation by 20% with less than 1% performance loss, demonstrating strong potential for efficient deployment of MoE-based models.

Data plays a pivotal role in the groundbreaking advancements in artificial intelligence. The quantitative analysis of data significantly contributes to model training, enhancing both the efficiency and quality of data utilization. However, existing data analysis tools often lag in accuracy. For instance, many of these tools even assume that the loss function of neural networks is convex. These limitations make it challenging to implement current methods effectively. In this paper, we introduce a new formulation to approximate a sample's influence by accumulating the differences in influence between consecutive learning steps, which we term Diff-In. Specifically, we formulate the sample-wise influence as the cumulative sum of its changes/differences across successive training iterations. By employing second-order approximations, we approximate these difference terms with high accuracy while eliminating the need for model convexity required by existing methods. Despite being a second-order method, Diff-In maintains computational complexity comparable to that of first-order methods and remains scalable. This efficiency is achieved by computing the product of the Hessian and gradient, which can be efficiently approximated using finite differences of first-order gradients. We assess the approximation accuracy of Diff-In both theoretically and empirically. Our theoretical analysis demonstrates that Diff-In achieves significantly lower approximation error compared to existing influence estimators. Extensive experiments further confirm its superior performance across multiple benchmark datasets in three data-centric tasks: data cleaning, data deletion, and coreset selection. Notably, our experiments on data pruning for large-scale vision-language pre-training show that Diff-In can scale to millions of data points and outperforms strong baselines.

Although processing-in-memory (PIM) techniques have widely been used for deep neural networks (DNNs) acceleration, the inference performance of aged PIM-based accelerators remains to be investigated. This paper makes the first attempt to study Hot Carrier Injection (HCI) and Negative Bias Temperature Instability (NBTI) aging impacts on SRAM-based DNN accelerators, which provides a novel and unified framework, termedAgeBalancefor aging detection, analysis and mitigation. First, we discuss a convenient aging detection scheme. Then, we benchmark the inference accuracy drops of DNNs running on aged SRAM-based PIM accelerators. Finally, we propose a low-cost anti-aging training method without incurring additional hardware overhead on SRAM-based DNN accelerators. Extensive experimental results on MNIST, CIFAR10 and AG News datasets show that aging can cause the inference accuracy of shallow or deep DNNs to drop to about 10%, close to random guessing. The aging mitigation scheme proposed in this paper can largely restore the accuracy to the original. Moreover, the SRAM write overhead of our method is much reduced thanks to a score-based training approach, leading to a reduction of 5× to 10× writing energy compared to the traditional training method.

Animatable 3D assets, defined as geometry equipped with an articulated skeleton and skinning weights, are fundamental to interactive graphics, embodied agents, and animation production. While recent 3D generative models can synthesize visually plausible shapes from images, the results are typically static. Obtaining usable rigs via post-hoc auto-rigging is brittle and often produces skeletons that are topologically inconsistent with the generated geometry. We present AniGen , a unified framework that directly generates animate-ready 3D assets conditioned on a single image. Our key insight is to represent shape, skeleton, and skinning as mutually consistent S 3 Fields (Shape, Skeleton, Skin) defined over a shared spatial domain. To enable the robust learning of these fields, we introduce two technical innovations: (i) a confidence-decaying skeleton field that explicitly handles the geometric ambiguity of bone prediction at Voronoi boundaries, and (ii) a dual skin feature field that decouples skinning weights from specific joint counts, allowing a fixed-architecture network to predict rigs of arbitrary complexity. Built upon a two-stage flow-matching pipeline, AniGen first synthesizes a sparse structural scaffold and then generates dense geometry and articulation in a structured latent space. Extensive experiments demonstrate that AniGen substantially outperforms state-of-the-art sequential baselines in rig validity and animation quality, generalizing effectively to in-the-wild images across diverse categories including animals, humanoids, and machinery. Homepage : https://yihua7.github.io/AniGen_web/

Pre-trained vision models (PVMs) are fundamental to modern robotics, yet their optimal configuration remains unclear. Through systematic evaluation, we find that while DINO and iBOT outperform MAE across visuomotor control and perception tasks, they struggle when trained on non-(single-)object-centric (NOC) data—a limitation strongly correlated with their diminished ability to learn object-centric representations. This investigation indicates that the ability to form object-centric representations from the non-object-centric robotics dataset is the key to success for PVMs. Motivated by this discovery, we designed SlotMIM, a method that induces object-centric representations by introducing a semantic bottleneck to reduce the number of prototypes to encourage the emergence of objectness as well as cross-view consistency regularization for encouraging multiview invariance. Our experiments encompass pre-training on object-centric, scene-centric, web-crawled, and ego-centric data. Across all settings, our approach learns transferrable representations and achieves significant improvements over prior work in image recognition, scene understanding, and robot learning evaluations. When scaled up with million-scale datasets, our method also demonstrates superior data efficiency and scalability. Our code and models are publicly available at https://github.com/CVMI-Lab/SlotMIM.

The advancement of Large Vision Language Models (LVLMs) has significantly improved multimodal understanding, yet challenges remain in video reasoning tasks due to the scarcity of high-quality, large-scale datasets. Existing video question-answering (VideoQA) datasets often rely on costly manual annotations with insufficient granularity or automatic construction methods with redundant frame-by-frame analysis, limiting their scalability and effectiveness for complex reasoning. To address these challenges, we introduce VideoEspresso, a novel dataset that features VideoQA pairs preserving essential spatial details and temporal coherence, along with multimodal annotations of intermediate reasoning steps. Our construction pipeline employs a semantic-aware method to reduce redundancy, followed by generating QA pairs using GPT-4o. We further develop video Chain-of-Thought (CoT) annotations to enrich reasoning processes, guiding GPT-4o in extracting logical relationships from QA pairs and video content. To exploit the potential of high-quality VideoQA pairs, we propose a Hybrid LVLMs Collaboration framework, featuring a Frame Selector and a two-stage instruction fine-tuned reasoning LVLM. This framework adaptively selects core frames and performs CoT reasoning using multimodal evidence. Evaluated on our proposed benchmark with 14 tasks against 9 popular LVLMs, our method outperforms existing baselines on most tasks, demonstrating superior video reasoning capabilities. Our code and dataset have been released at: https://github.com/hshjerry/VideoEspresso

Generating high-fidelity, controllable, and annotated training data is critical for autonomous driving. Existing methods typically generate a single data form directly from a coarse scene layout, which not only fails to output rich data forms required for diverse downstream tasks but also struggles to model the direct layout-to-data distribution. In this paper, we introduce UniScene, the first unified framework for generating three key data forms — semantic occupancy, video, and LiDAR — in driving scenes. UniScene employs a progressive generation process that decomposes the complex task of scene generation into two hierarchical steps: (a) first generating semantic occupancy from a customized scene layout as a meta scene representation rich in both semantic and geometric information, and then (b) conditioned on occupancy, generating video and LiDAR data, respectively, with two novel transfer strategies of Gaussian-based Joint Rendering and Prior-guided Sparse Modeling. This occupancy-centric approach reduces the generation burden, especially for intricate scenes, while providing detailed intermediate representations for the subsequent generation stages. Extensive experiments demonstrate that UniScene outperforms previous SOTAs in the occupancy, video, and LiDAR generation, which also indeed benefits downstream driving tasks. The Project is available at https://arlo0o.github.io/uniscene/.

Simple as it seems, moving an object to another location within an image is, in fact, a challenging image-editing task that requires re-harmonizing the lighting, adjusting the pose based on perspective, accurately filling occluded regions, and ensuring coherent synchronization of shadows and reflections while maintaining the object identity. In this paper, we present ObjectMover, a generative model that can perform object movement in highly challenging scenes. Our key insight is that we model this task as a sequence-to-sequence problem and fine-tune a video generation model to leverage its knowledge of consistent object generation across video frames. We show that with this approach, our model is able to adjust to complex real-world scenarios, handling extreme lighting harmonization and object effect movement. As large-scale data for object movement are unavailable, we construct a data generation pipeline using a modern game engine to synthesize high-quality data pairs. We further propose a multi-task learning strategy that enables training on real-world video data to improve the model generalization. Through extensive experiments, we demonstrate that ObjectMover achieves outstanding results and adapts well to real-world scenarios.

Balancing training on long-tail data distributions remains a long-standing challenge in deep learning. While methods such as re-weighting and re-sampling help alleviate the imbalance issue, limited sample diversity continues to hinder models from learning robust and generalizable feature representations, particularly for tail classes. In contrast to existing methods, we offer a novel perspective on long-tail learning, inspired by an observation: datasets with finer granularity tend to be less affected by data imbalance. In this paper, we investigate this phenomenon through both quantitative and qualitative studies, showing that increased granularity enhances the generalization of learned features in tail categories. Motivated by these findings, we propose a method to increase dataset granularity through category extrapolation. Specifically, we introduce open-set fine-grained classes that are related to existing ones, aiming to enhance representation learning for both head and tail classes. To automate the curation of auxiliary data, we leverage large language models (LLMs) as knowledge bases to search for auxiliary categories and retrieve relevant images through web crawling. To prevent the overwhelming presence of auxiliary classes from disrupting training, we introduce a neighbor-silencing loss that encourages the model to focus on class discrimination within the target dataset. During inference, the classifier weights for auxiliary categories are masked out, leaving only the target class weights for use. Extensive experiments on three standard long-tail benchmarks demonstrate the effectiveness of our approach, notably outperforming strong baseline methods that use the same amount of data. The project is available at shizhen-zhao.github.io/LT_Neighbors/.

Vision Transformers (ViTs) are new foundation models for vision applications. Edge-deploying ViTs to realize energy-saving, low-latency, and high-performance dense predictions have wide applications, such as autonomous driving and surveillance image analysis. However, the quadratic complexity of the self-attention mechanism renders ViTs slow and resource-intensive, particularly for pixel-level dense predictions that involve long contexts. Additionally, the pyramid-like architecture of modern ViT variants leads to an unbalanced workload, further reducing hardware utilization and decreasing the throughput of conventional edge devices. To this end, we propose an algorithm-hardware co-optimized edge ViT accelerator tailored for efficient dense predictions. At the algorithm level, we propose a decoupled chunk attention (DCA) mechanism implemented in a pipelined manner to reduce off-chip memory access, thereby enabling efficient dense predictions within limited on-chip memory. At the architecture level, we introduce a hybrid architecture that combines SRAM-based computing-in-memory (CIM) and nonvolatile RRAM storage to eliminate extensive off-chip memory access, with a fusion scheduling to balance workloads and minimize intermediate on-chip memory access. At the circuit level, a bit/element two-way-reconfigurable CIM macro is proposed to improve hardware utilization across pyramidal ViT blocks with varied matrix sizes. The experimental results on object detection, semantic segmentation, and depth estimation tasks demonstrate that our design can efficiently process patch lengths up to 16384 with a speedup of 18.5×-217.1×, a reduction in memory accesses of 1.7×-7.4×, and an improvement in energy efficiency of 1.8×, under less than 1% performance degradation.

Deploying deep neural networks (DNNs) on conventional digital edge devices faces significant challenges due to high energy consumption. A promising solution is the processing-inmemory (PIM) architecture with resistive random-access memory (RRAM), but RRAM-based systems suffer from imprecise weights due to programming stochasticity and cannot effectively utilize conventional weight encryption/decryption intellectual property (IP) protection schemes. To address these issues, we propose a novel software-hardware co-design Guarder. On the hardware side, we introduce 3T2R cells to achieve reliable multiply-accumulate (MAC) operations and use reconfigurable inverter operating voltages to encode keys for encrypting DNNs on RRAM. On the software side, we implement a contrastive training method that ensures high model accuracy on authorized chips while degrading performance on unauthorized ones. This approach protects DNN IP with minimal hardware overhead while significantly mitigating the effects of RRAM programming stochasticity. Extensive experiments on tasks such as image classification (using MLP, ResNet, and ViT), segmentation (using SegFormer), and image generation (using DiT) validate the effectiveness of our method. The proposed contrastive training ensures negligible performance degradation on authorized chips, while performance on unauthorized chips drops to random guessing or generation. Compared to traditional RRAM accelerators, the 3T2R-based accelerator achieves a $1.41 \times$ reduction in area overhead and a $2.28 \times$ reduction in energy consumption.

Understanding abnormal events in videos is a vital and challenging task that has garnered significant attention in a wide range of applications. Although current video understanding Multi-modal Large Language Models (MLLMs) are capable of analyzing general videos, they often struggle to handle anomalies due to the spatial and temporal sparsity of abnormal events, where the redundant information always leads to suboptimal outcomes. To address these challenges, exploiting the representation and generalization capabilities of Vison Language Models (VLMs) and Large Language Models (LLMs), we propose VA-GPT, a novel MLLM designed for summarizing and localizing abnormal events in various videos. Our approach efficiently aligns effective tokens between visual encoders and LLMs through two key proposed modules: Spatial Effective Token Selection (SETS) and Temporal Effective Token Generation (TETG). These modules enable our model to effectively capture and analyze both spatial and temporal information associated with abnormal events, resulting in more accurate responses and interactions. Furthermore, we construct an instruction-following dataset specifically for fine-tuning video-anomaly-aware MLLMs, and introduce a cross-domain evaluation benchmark based on XD-Violence dataset. Our proposed method outperforms existing state-of-the-art methods on various benchmarks.

Current generative models struggle to synthesize dynamic 4D driving scenes that simultaneously support temporal extrapolation and spatial novel view synthesis (NVS) without per-scene optimization. A key challenge lies in finding an efficient and generalizable geometric representation that seamlessly connects temporal and spatial synthesis. To address this, we propose DiST-4D, the first disentangled spatiotemporal diffusion framework for 4D driving scene generation, which leverages metric depth as the core geometric representation. DiST-4D decomposes the problem into two diffusion processes: DiST-T, which predicts future metric depth and multi-view RGB sequences directly from past observations, and DiST-S, which enables spatial NVS by training only on existing viewpoints while enforcing cycle consistency. This cycle consistency mechanism introduces a forward-backward rendering constraint, reducing the generalization gap between observed and unseen viewpoints. Metric depth is essential for both accurate reliable forecasting and accurate spatial NVS, as it provides a view-consistent geometric representation that generalizes well to unseen perspectives. Experiments demonstrate that DiST-4D achieves state-of-the-art performance in both temporal prediction and NVS tasks, while also delivering competitive performance in planning-related evaluations.

Video generation models have demonstrated great capability of producing impressive monocular videos, however, the generation of 3D stereoscopic video remains under-explored. We propose a pose-free and training-free approach for generating 3D stereoscopic videos using an off-the-shelf monocular video generation model. Our method warps a generated monocular video into camera views on stereoscopic baseline using estimated video depth, and employs a novel frame matrix video inpainting framework. The framework leverages the video generation model to inpaint frames observed from different timestamps and views. This effective approach generates consistent and semantically coherent stereoscopic videos without scene optimization or model fine-tuning. Moreover, we develop a disocclusion boundary re-injection scheme that further improves the quality of video inpainting by alleviating the negative effects propagated from disoccluded areas in the latent space. We validate the efficacy of our proposed method by conducting experiments on videos from various generative models, including Sora [4], Lumiere [2], WALT [8], and Zeroscope [12]. The experiments demonstrate that our method has a significant improvement over previous methods. Project page at https://daipengwa.github.io/SVG_ProjectPage/

Mixture-of-Experts large language models (MoE-LLMs) marks a significant step forward of language models, however, they encounter two critical challenges in practice: 1) expert parameters lead to considerable memory consumption and loading latency; and 2) the current activated experts are redundant, as many tokens may only require a single expert. Motivated by these issues, we investigate the MoE-LLMs and make two key observations: a) different experts exhibit varying behaviors on activation reconstruction error, routing scores, and activated frequencies, highlighting their differing importance, and b) not all tokens are equally important-- only a small subset is critical. Building on these insights, we propose MC, a training-free Mixture-Compressor for MoE-LLMs, which leverages the significance of both experts and tokens to achieve an extreme compression. First, to mitigate storage and loading overheads, we introduce Pre-Loading Mixed-Precision Quantization (PMQ), which formulates the adaptive bit-width allocation as a Linear Programming (LP) problem, where the objective function balances multi-factors reflecting the importance of each expert. Additionally, we develop Online Dynamic Pruning (ODP), which identifies important tokens to retain and dynamically select activated experts for other tokens during inference to optimize efficiency while maintaining performance. Our MC integrates static quantization and dynamic pruning to collaboratively achieve extreme compression for MoE-LLMs with less accuracy loss, ensuring an optimal trade-off between performance and efficiency Extensive experiments confirm the effectiveness of our approach. For instance, at 2.54 bits, MC compresses 76.6% of the model, with only a 3.8% average accuracy loss. During dynamic inference, we further reduce activated parameters by 15%, with a performance drop of less than 0.6%. Remarkably, MC even surpasses floating-point 13b dense LLMs with significantly smaller parameter sizes, suggesting that mixture compression in MoE-LLMs has the potential to outperform both comparable and larger dense LLMs. Our code is available at https://github.com/Aaronhuang-778/MC-MoE

In this paper, we present InfoMax, a novel data pruning method, also known as coreset selection, designed to maximize the information content of selected samples while minimizing redundancy. By doing so, InfoMax enhances the overall informativeness of the coreset. The information of individual samples is measured by importance scores, which capture their influence or difficulty in model learning. To quantify redundancy, we use pairwise sample similarities, based on the premise that similar samples contribute similarly to the learning process. We formalize the coreset selection problem as a discrete quadratic programming (DQP) task, with the objective of maximizing the total information content, represented as the sum of individual sample contributions minus the redundancies introduced by similar samples within the coreset. To ensure practical scalability, we introduce an efficient gradient-based solver, complemented by sparsification techniques applied to the similarity matrix and dataset partitioning strategies. This enables InfoMax to seamlessly scale to datasets with millions of samples. Extensive experiments demonstrate the superior performance of InfoMax in various data pruning tasks, including image classification, vision-language pre-training, and instruction tuning for large language models.

Post-training quantization (PTQ) is an effective technique for compressing large language models (LLMs). However, while uniform-precision quantization is computationally efficient, it often compromises model performance. To address this, we propose SliM-LLM, a salience-driven mixed-precision quantization framework that allocates bit-widths at the group-wise with high accuracy. Our approach leverages the observation that important weights follow a structured distribution and introduces two key components: 1) Salience-Determined Bit Allocation adaptively assigns bit-widths to groups within each layer based on their salience; and 2) Salience-Weighted Quantizer Calibration optimizes quantizer parameters by incorporating element-level salience, retain essential information. With its structured group-wise partitioning, SliM-LLM provides a hardware-friendly solution that matches the efficiency of uniform quantization methods while significantly improving accuracy. Experiments show that SliM-LLM achieves superior performance across various LLMs at low bit-widths. For example, a 2-bit quantized LLaMA-7B model reduces memory usage by nearly 6x compared to the floating-point baseline, decreases perplexity by 48% compared to state-of-the-art gradient-free PTQ methods, and maintains GPU inference speed. Additionally, the extended version, SliM-LLM+, which incorporates gradient-based quantization, further reduces perplexity by 35.1%. Our code is available at https://github.com/Aaronhuang-778/SliM-LLM.

We introduce a full-stack framework that scales up reasoning in vision-language models (VLMs) to long videos, leveraging reinforcement learning. We address the unique challenges of long video reasoning by integrating three critical components: (1) a large-scale dataset, LongVideo-Reason, comprising 104K long video QA pairs with high-quality reasoning annotations across diverse domains such as sports, games, and vlogs; (2) a two-stage training pipeline that extends VLMs with chain-of-thought supervised fine-tuning (CoT-SFT) and reinforcement learning (RL); and (3) a training infrastructure for long video RL, named Multi-modal Reinforcement Sequence Parallelism (MR-SP), which incorporates sequence parallelism and a vLLM-based engine tailored for long video, using cached video embeddings for efficient rollout and prefilling. In our experiments, LongVILA-R1-7B achieves strong performance on video benchmarks, reaching 65.1% and 71.1% accuracy on VideoMME without and with subtitles, respectively, and consistently outperforming LongVILA-7B across multiple benchmarks. Moreover, LongVILA-R1-7B supports processing up to 8,192 video frames per video, and configurable FPS settings. Notably, our MR-SP system achieves up to 2.1x speedup on long video RL training. In addition, we release our training system for public availability that supports RL training on various modalities (video, text, and audio), various models (VILA and Qwen series), and even image and video generation models. On a single A100 node (8 GPUs), it supports RL training on hour-long videos (e.g., 3,600 frames). Code and models are available at https://github.com/NVlabs/Long-RL

Visual generative and understanding models typically rely on distinct tokenizers to process images, presenting a key challenge for unifying them within a single framework. Recent studies attempt to address this by connecting the training of VQVAE (for autoregressive generation) and CLIP (for understanding) to build a unified tokenizer. However, directly combining these training objectives has been observed to cause severe loss conflicts. In this paper, we show that reconstruction and semantic supervision do not inherently conflict. Instead, the underlying bottleneck stems from limited representational capacity of discrete token space. Building on these insights, we introduce UniTok, a unified tokenizer featuring a novel multi-codebook quantization mechanism that effectively scales up the vocabulary size and bottleneck dimension. In terms of final performance, UniTok sets a new record of 0.38 rFID and 78.6\% zero-shot accuracy on ImageNet. Besides, UniTok can be seamlessly integrated into MLLMs to unlock native visual generation capability, without compromising the understanding performance. Additionally, we show that UniTok favors cfg-free generation, reducing gFID from 14.6 to 2.5 on ImageNet 256$\times$256 benchmark. All codes and models have been made publicly available.

Reinforcement fine-tuning (RFT) is essential for enhancing the reasoning and generalization capabilities of large language models, but its success heavily relies on the quality of the training data. While data selection has been extensively studied in supervised learning, its role in reinforcement learning, particularly during the RFT stage, remains largely underexplored. In this work, we introduce RFT-Inf, the first influence estimator designed for data in reinforcement learning. RFT-Inf quantifies the importance of each training example by measuring how its removal affects the final training reward, offering a direct estimate of its contribution to model learning. To ensure scalability, we propose a first-order approximation of the RFT-Inf score by backtracking through the optimization process and applying temporal differentiation to the sample-wise influence term, along with a first-order Taylor approximation to adjacent time steps. This yields a lightweight, gradient-based estimator that evaluates the alignment between an individual sample’s gradient and the average gradient direction of all training samples, where a higher degree of alignment implies greater training utility. Extensive experiments demonstrate that RFT-Inf consistently improves reward performance and accelerates convergence in reinforcement fine-tuning.

Long chain-of-thought (LongCoT) has emerged as a powerful reasoning paradigm for enabling large language models (LLMs) to solve complex tasks through a systematic and thorough thinking phase. Although supervised fine-tuning (SFT) on high-quality LongCoT traces has proven effective to activate LongCoT abilities, we find that models trained in this way tend to overfit problem-specific knowledge and heuristics, leading to degraded out-of-distribution performance. To address this issue, we propose a Decoupled LongCoT Fine-Tuning (DLoFT) algorithm, which enables the model to learn generalizable LongCoT reasoning abilities while preventing overfitting to the reasoning content with problem-specific information. The key idea is to decouple the gradient into two orthogonal components: 1) a paradigm-relevant gradient corresponding to the general LongCoT paradigm and 2) a content-relevant gradient reflecting the problem-specific information, where only the former gradient is used to update model parameters. Specifically, by leveraging the unique two-phase composition (thinking and solution) of the LongCoT response, our gradient decoupling mechanism isolates the content-relevant gradient via a projection operation and separates the paradigm-relevant gradient through orthogonalization. Our DLoFT ensures the model concentrate on internalizing the LongCoT paradigm rather than memorizing problem-specific knowledge and heuristics. Extensive experiments demonstrate that our DLoFT significantly improves the generalization behavior of LongCoT abilities compared to SFT while maintaining strong in-distribution performance.

Robust 3D geometry estimation from videos is critical for applications such as autonomous navigation, SLAM, and 3D scene reconstruction. Recent methods like DUSt3R demonstrate that regressing dense pointmaps from image pairs enables accurate and efficient pose-free reconstruction. However, existing RGB-only approaches struggle under real-world conditions involving dynamic objects and extreme illumination, due to the inherent limitations of conventional cameras. In this paper, we propose \textbf{EAG3R}, a novel geometry estimation framework that augments pointmap-based reconstruction with asynchronous event streams. Built upon the MonST3R backbone, EAG3R introduces two key innovations: (1) a retinex-inspired image enhancement module and a lightweight event adapter with SNR-aware fusion mechanism that adaptively combines RGB and event features based on local reliability; and (2) a novel event-based photometric consistency loss that reinforces spatiotemporal coherence during global optimization. Our method enables robust geometry estimation in challenging dynamic low-light scenes without requiring retraining on night-time data. Extensive experiments demonstrate that EAG3R significantly outperforms state-of-the-art RGB-only baselines across monocular depth estimation, camera pose tracking, and dynamic reconstruction tasks.

Recent text-to-image systems face limitations in handling multimodal inputs and complex reasoning tasks. We introduce MindOmni, a unified multimodal large language model that addresses these challenges by incorporating reasoning generation through reinforcement learning. MindOmni leverages a three-phase training strategy: i) design of a unified vision language model with a decoder-only diffusion module, ii) supervised fine-tuning with Chain-of-Thought (CoT) instruction data, and iii) our proposed Reasoning Generation Policy Optimization (RGPO) algorithm, utilizing multimodal feedback to effectively guide policy updates. Experimental results demonstrate that MindOmni outperforms existing models, achieving impressive performance on both understanding and generation benchmarks, meanwhile showcasing advanced fine-grained reasoning generation capabilities, especially with mathematical reasoning instruction. All codes will be made public.

In this work, we present a novel direction to build an image tokenizer directly on top of a frozen vision foundation model, which is a largely underexplored area. Specifically, we employ a frozen vision foundation model as the encoder of our tokenizer. To enhance its effectiveness, we introduce two key components: (1) a region-adaptive quantization framework that reduces redundancy in the pre-trained features on regular 2D grids, and (2) a semantic reconstruction objective that aligns the tokenizer’s outputs with the foundation model’s representations to preserve semantic fidelity. Based on these designs, our proposed image tokenizer, \textbf{\ours}, achieves substantial improvements in image reconstruction and generation quality, while also enhancing token efficiency. It further boosts autoregressive (AR) generation---achieving a gFID of \textbf{1.36} on ImageNet benchmarks, while accelerating model convergence by \textbf{three times}, and enabling high-fidelity class-conditional synthesis without the need for classifier-free guidance (CFG). The code is available at \href{https://github.com/CVMI-Lab/VFMTok}{https://github.com/CVMI-Lab/VFMTok}.

Training native 3D texture generative models remains a fundamental yet challenging problem, largely due to the limited availability of large-scale, high-quality 3D texture datasets. This scarcity hinders generalization to real-world scenarios. To address this, most existing methods finetune foundation image generative models to exploit their learned visual priors. However, these approaches typically generate only multi-view images and rely on post-processing to produce UV texture maps—an essential representation in modern graphics pipelines. Such two-stage pipelines often suffer from error accumulation and spatial inconsistencies across the 3D surface. In this paper, we introduce SeqTex, a novel end-to-end framework that leverages the visual knowledge encoded in pretrained video foundation models to directly generate complete UV texture maps. Unlike previous methods that model the distribution of UV textures in isolation, SeqTex reformulates the task as a sequence generation problem, enabling the model to learn the joint distribution of multi-view renderings and UV textures. This design effectively transfers the consistent image-space priors from video foundation models into the UV domain. To further enhance performance, we propose several architectural innovations: a decoupled multi-view and UV branch design, geometry-informed attention to guide cross-domain feature alignment, and adaptive token resolution to preserve fine texture details while maintaining computational efficiency. Together, these components allow SeqTex to fully utilize pretrained video priors and synthesize high-fidelity UV texture maps without the need for post-processing. Extensive experiments show that SeqTex achieves state-of-the-art performance on both image-conditioned and text-conditioned 3D texture generation tasks, with superior 3D consistency, texture-geometry alignment, and real-world generalization. Our project page is https://yuanze1024.github.io/SeqTex/.

Users often take notes for instructional videos to access key knowledge later without revisiting long videos.Automated note generation tools enable users to obtain informative notes efficiently.However, notes generated by existing research or off-the-shelf tools fail to preserve the information conveyed in the original videos comprehensively, nor can they satisfy users' expectations for diverse presentation formats and interactive features when using notes digitally.In this work, we present NoteIt, a system, which automatically converts instructional videos to interactable notes using a novel pipeline that faithfully extracts hierarchical structure and multimodal key information from videos.With NoteIt's interface, users can interact with the system to further customize the content and presentation formats of the notes according to their preferences.We conducted both a technical evaluation and a comparison user study (N=36).The solid performance in objective metrics and the positive user feedback demonstrated the effectiveness of the pipeline and the overall usability of NoteIt.

Virtual environments (VEs) are pivotal for virtual, augmented, and mixed reality systems. Despite advances in 3D generation and reconstruction, the direct creation of 3D objects within an established 3D scene (represented as NeRF) for novel VE creation remains a relatively unexplored domain. This process is complex, requiring not only the generation of high-quality 3D objects but also their seamless integration into the existing scene. To this end, we propose a novel pipeline featuring an intuitive interface, dubbed GO-NeRF. Our approach takes text prompts and user-specified regions as inputs and leverages the scene context to generate 3D objects within the scene. We employ a compositional rendering formulation that effectively integrates the generated 3D objects into the scene, utilizing optimized 3D-aware opacity maps to avoid unintended modifications to the original scene. Furthermore, we develop tailored optimization objectives and training strategies to enhance the model's ability to capture scene context and mitigate artifacts, such as floaters, that may occur while optimizing 3D objects within the scene. Extensive experiments conducted on both forward-facing and 360°scenes demonstrate the superior performance of our proposed method in generating objects that harmonize with surrounding scenes and synthesizing high-quality novel view images. The code will be at https://daipengwa.github.io/G0-NeRF/.

Dominant dual-encoder models enable efficient image-text retrieval but suffer from limited accuracy, while the cross-encoder models offer higher accuracy at the expense of efficiency. Distilling cross-modality matching knowledge from cross-encoder to dual-encoder provides a natural approach to harness their strengths. Thus, we investigate the following valuable question: how to make cross-encoder a good teacher for dual-encoder? Our findings are threefold: (1) Cross-modal similarity score distribution of cross-encoder is more concentrated, while the result of dual-encoder is nearly normal, making vanilla logit distillation less effective. However, ranking distillation remains practical, as it is not affected by the score distribution. (2) Only the relative order between hard negatives conveys valid knowledge, while the order information between easy negatives has little significance. (3) Maintaining the coordination between distillation loss and dual-encoder training loss is beneficial for knowledge transfer. Based on these findings, we propose a novel Contrastive Partial Ranking Distillation (CPRD) method, which implements the objective of mimicking relative order between hard negative samples with contrastive learning. This approach coordinates with the training of the dual-encoder, effectively transferring valid knowledge from the cross-encoder to the dual-encoder. Extensive experiments on image-text retrieval and ranking tasks show that our method surpasses other distillation methods and significantly improves the accuracy of dual-encoder.

In this paper, we present an empirical study on image recognition unfairness, i.e., extreme class accuracy disparity on balanced data like ImageNet. We demonstrate that classes are not equal and unfairness is prevalent for image classification models across various datasets, network architectures, and model capacities. Moreover, several intriguing properties of fairness are identified. First, the unfairness lies in problematic representation rather than classifier bias distinguished from long-tailed recognition. Second, with the proposed concept of Model Prediction Bias, we investigate the origins of problematic representation during training optimization. Our findings reveal that models tend to exhibit greater prediction biases for classes that are more challenging to recognize. It means that more other classes will be confused with harder classes. Then the False Positives (FPs) will dominate the learning in optimization, thus leading to their poor accuracy. Further, we conclude that data augmentation and representation learning algorithms improve overall performance by promoting fairness to some degree in image classification.

Novel view synthesis for dynamic scenes is still a challenging problem in computer vision and graphics. Recently, Gaussian splatting has emerged as a robust technique to represent static scenes and enable high-quality and real-time novel view synthesis. Building upon this technique, we propose a new representation that explicitly decomposes the motion and appearance of dynamic scenes into sparse control points and dense Gaussians, respectively. Our key idea is to use sparse control points, significantly fewer in number than the Gaussians, to learn compact 6 DoF transformation bases, which can be locally interpolated through learned interpolation weights to yield the motion field of 3D Gaussians. We employ a deformation MLP to predict time-varying 6 DoF transformations for each control point, which reduces learning complexities, enhances learning abilities, and facilitates obtaining temporal and spatial coherent motion patterns. Then, we jointly learn the 3D Gaussians, the canonical space locations of control points, and the deformation MLP to reconstruct the appearance, geometry, and dynamics of 3D scenes. During learning, the location and number of control points are adaptively adjusted to accommodate varying motion complexities in different regions, and an ARAP loss following the principle of as rigid as possible is developed to enforce spatial continuity and local rigidity of learned motions. Finally, thanks to the explicit sparse motion representation and its decomposition from appearance, our method can enable user-controlled motion editing while retaining high-fidelity appearances. Extensive experiments demonstrate that our approach outperforms existing approaches on novel view synthesis with a high rendering speed and enables novel appearance-preserved motion editing applications.

We introduce EscherNet, a multi-view conditioned diffusion model for view synthesis. EscherNet learns implicit and generative 3D representations coupled with a specialised camera positional encoding, allowing precise and continuous relative control of the camera transformation between an arbitrary number of reference and target views. EscherNet offers exceptional generality, flexibility, and scalability in view synthesis ─it can generate more than 100 consistent target views simultaneously on a single consumer-grade GPU, despite being trained with a fixed number of 3 reference views to 3 target views. As a result, EscherNet not only addresses zero-shot novel view synthesis, but also naturally unifies single- and multi-image 3D reconstruction, combining these diverse tasks into a single, cohesive framework. Our extensive experiments demonstrate that EscherNet achieves state-of-the-art performance in multiple benchmarks, even when compared to methods specifically tailored for each individual problem. This remarkable versatility opens up new directions for designing scalable neural architectures for 3D vision. Project page: https://kxhit.github.io/EscherNet.

Scene reconstruction from multi-view images is a fun- damental problem in computer vision and graphics. Re- cent neural implicit surface reconstruction methods have achieved high-quality results; however, editing and ma- nipulating the 3D geometry of reconstructed scenes re- mains challenging due to the absence of naturally decom- posed object entities and complex object/background com- positions. In this paper, we present Total-Decom, a novel method for decomposed 3D reconstruction with minimal human interaction. Our approach seamlessly integrates the Segment Anything Model (SAM) with hybrid implicit- explicit neural surface representations and a mesh-based region-growing technique for accurate 3D object decom- position. Total-Decom requires minimal human annotations while providing users with real-time control over the granularity and quality of decomposition. We extensively evaluate our method on benchmark datasets and demon- strate its potential for downstream applications, such as animation and scene editing. The code is available at https://github.com/CVMI-Lab/Total-Decom.git.

Vision-language pre-training (VLP) aims to learn joint representations of vision and language modalities. The contrastive paradigm is currently dominant in this field. However, we observe a notable misalignment phenomenon, that is, the affinity between samples has an obvious disparity across different modalities, namely “Affinity Inconsistency Problem”. Our intuition is that, for a well-aligned model, two images that look similar to each other should have the same level of similarity as their corresponding texts that describe them. In this paper, we first investigate the reason of this inconsistency problem. We discover that the lack of consideration for sample-wise affinity consistency across modalities in existing training objectives is the central cause. To address this problem, we propose a novel loss function, named Sample-wise affinity Consistency (SaCo) loss, which is designed to enhance such consistency by minimizing the distance between image embedding similarity and text embedding similarity for any two samples. Our SaCo loss can be easily incorporated into existing vision-language models as an additional loss due to its complementarity for most training objectives. In addition, considering that pre-training from scratch is computationally expensive, we also provide a more efficient way to continuously pre-train on a converged model by integrating our loss. Experimentally, the model trained with our SaCo loss significantly outperforms the baseline on a variety of vision and language tasks.

We propose a lightweight and scalable Regional Point-Language Contrastive learning framework, namely RegionPLC, for open-world 3D scene understanding, aiming to identify and recognize open-set objects and categories. Specifically, based on our empirical studies, we introduce a 3D-aware SFusion strategy that fuses 3D vision-language pairs derived from multiple 2D foundation models, yielding high-quality, dense region-level language descriptions without human 3D annotations. Subsequently, we devise a region-aware point-discriminative contrastive learning objective to enable robust and effective 3D learning from dense regional language supervision. We carry out extensive experiments on ScanNet, ScanNet200, and nuScenes datasets, and our model outperforms prior 3D open-world scene understanding approaches by an average of 17.2% and 9.1% for semantic and instance segmentation, respectively, while maintaining greater scalability and lower resource demands. Furthermore, our method has the flexibility to be effortlessly integrated with language models to enable open-ended grounded 3D reasoning without extra task-specific training. Code will be released at github.

Biologically plausible Spiking Neural Networks (SNNs), characterized by spike sparsity, are growing tremendous attention over intellectual edge devices and critical bio-medical applications as compared to artificial neural networks (ANNs). However, there is a considerable risk from malicious attempts to extract white-box information (i.e., weights) from SNNs, as attackers could exploit well-trained SNNs for profit and white-box adversarial concerns. There is a dire need for intellectual property (IP) protective measures. In this paper, we present a novel secure software-hardware co-designed RRAM-based neuromorphic accelerator for protecting the IP of SNNs. Software-wise, we design a tailored genetic algorithm with classic XOR encryption to target the least number of weights that need encryption. From a hardware perspective, we develop a low-energy decryption module, meticulously designed to provide zero decryption latency. Extensive results from various datasets, including NMNIST, DVSGesture, EEGMMIDB, Braille Letter, and SHD, demonstrate that our proposed method effectively secures SNNs by encrypting a minimal fraction of stealthy weights, only 0.00005% to 0.016% weight bits. Additionally, it achieves a substantial reduction in energy consumption, ranging from ×59 to ×6780, and significantly lowers decryption latency, ranging from ×175 to ×4250. Moreover, our method requires as little as one sample per class in dataset for encryption and addresses hessian/gradient-based search insensitive problems. This strategy offers a highly efficient and flexible solution for securing SNNs in diverse applications1.

Text-to-3D generation has made remarkable progress recently, particularly with methods based on Score Distillation Sampling (SDS) that leverages pre-trained 2D diffusion models. While the usage of classifier-free guidance is well acknowledged to be crucial for successful optimization, it is considered an auxiliary trick rather than the most essential component. In this paper, we re-evaluate the role of classifier-free guidance in score distillation and discover a surprising finding: the guidance alone is enough for effective text-to-3D generation tasks. We name this method Classifier Score Distillation (CSD), which can be interpreted as using an implicit classification model for generation. This new perspective reveals new insights for understanding existing techniques. We validate the effectiveness of CSD across a variety of text-to-3D tasks including shape generation, texture synthesis, and shape editing, achieving results superior to those of state-of-the-art methods. Our project page is https://xinyu-andy.github.io/Classifier-Score-Distillation

Pretrained large language models (LLMs) exhibit exceptional general language processing capabilities but come with significant demands on memory and computational resources. As a powerful compression technology, binarization can extremely reduce model weights to a mere 1 bit, lowering the expensive computation and memory requirements. However, existing quantization techniques fall short of maintaining LLM performance under ultra-low bit-widths. In response to this challenge, we present BiLLM, a groundbreaking 1-bit post-training quantization scheme tailored for pretrained LLMs. Based on the weight distribution of LLMs, BiLLM first identifies and structurally selects salient weights, and minimizes the compression loss through an effective binary residual approximation strategy. Moreover, considering the bell-shaped distribution of the non-salient weights, we propose an optimal splitting search to group and binarize them accurately. BiLLM, for the first time, achieves high-accuracy inference (e.g. 8.41 perplexity on LLaMA2-70B) with only 1.08-bit weights across various LLM families and evaluation metrics, outperforms SOTA quantization methods of LLM by significant margins. Moreover, BiLLM enables the binarization process of a 7-billion LLM within 0.5 hours on a single GPU, demonstrating satisfactory time efficiency. Our code is available at https://github.com/Aaronhuang-778/BiLLM .

Severe data imbalance naturally exists among web-scale vision-language datasets. Despite this, we find CLIP pre-trained thereupon exhibits notable robustness to the data imbalance compared to supervised learning, and demonstrates significant effectiveness in learning generalizable representations. With an aim to investigate the reasons behind this finding, we conduct controlled experiments to study various underlying factors, and reveal that CLIP's pretext task forms a dynamic classification problem wherein only a subset of classes is present in training. This isolates the bias from dominant classes and implicitly balances the learning signal. Furthermore, the robustness and discriminability of CLIP improve with more descriptive language supervision, larger data scale, and broader open-world concepts, which are inaccessible to supervised learning. Our study not only uncovers the mechanisms behind CLIP's generalizability beyond data imbalance but also provides transferable insights for the research community. The findings are validated in both supervised and self-supervised learning, enabling models trained on imbalanced data to achieve CLIP-level performance on diverse recognition tasks. Code and data are available at: https://github.com/CVMI-Lab/clip-beyond-tail.

In this paper, we delve deeper into the Kullback–Leibler (KL) Divergence loss and mathematically prove that it is equivalent to the Decoupled Kullback-Leibler (DKL) Divergence loss that consists of 1) a weighted Mean Square Error ($\mathbf{w}$MSE) loss and 2) a Cross-Entropy loss incorporating soft labels. Thanks to the decomposed formulation of DKL loss, we have identified two areas for improvement. Firstly, we address the limitation of KL/DKL in scenarios like knowledge distillation by breaking its asymmetric optimization property. This modification ensures that the $\mathbf{w}$MSE component is always effective during training, providing extra constructive cues. Secondly, we introduce class-wise global information into KL/DKL to mitigate bias from individual samples. With these two enhancements, we derive the Improved Kullback–Leibler (IKL) Divergence loss and evaluate its effectiveness by conducting experiments on CIFAR-10/100 and ImageNet datasets, focusing on adversarial training, and knowledge distillation tasks. The proposed approach achieves new state-of-the-art adversarial robustness on the public leaderboard --- \textit{RobustBench} and competitive performance on knowledge distillation, demonstrating the substantial practical merits. Our code is available at https://github.com/jiequancui/DKL.

Video representation is a long-standing problem that is crucial for various downstream tasks, such as tracking, depth prediction, segmentation, view synthesis, and editing. However, current methods either struggle to model complex motions due to the absence of 3D structure or rely on implicit 3D representations that are ill-suited for manipulation tasks. To address these challenges, we introduce a novel explicit 3D representation—video Gaussian representation—that embeds a video into 3D Gaussians. Our proposed representation models video appearance in a 3D canonical space using explicit Gaussians as proxies and associates each Gaussian with 3D motions for video motion. This approach offers a more intrinsic and explicit representation than layered atlas or volumetric pixel matrices. To obtain such a representation, we distill 2D priors, such as optical flow and depth, from foundation models to regularize learning in this ill-posed setting. Extensive applications demonstrate the versatility of our new video representation. It has been proven effective in numerous video processing tasks, including tracking, consistent video depth and feature refinement, motion and appearance editing, and stereoscopic video generation.

The recent advancements in 3D Gaussian splatting (3D-GS) have not only facilitated real-time rendering through modern GPU rasterization pipelines but have also attained state-of-the-art rendering quality. Nevertheless, despite its exceptional rendering quality and performance on standard datasets, 3D-GS frequently encounters difficulties in accurately modeling specular and anisotropic components. This issue stems from the limited ability of spherical harmonics (SH) to represent high-frequency information. To overcome this challenge, we introduce Spec-Gaussian, an approach that utilizes an anisotropic spherical Gaussian (ASG) appearance field instead of SH for modeling the view-dependent appearance of each 3D Gaussian. Additionally, we have developed a coarse-to-fine training strategy to improve learning efficiency and eliminate floaters caused by overfitting in real-world scenes. Our experimental results demonstrate that our method surpasses existing approaches in terms of rendering quality. Thanks to ASG, we have significantly improved the ability of 3D-GS to model scenes with specular and anisotropic components without increasing the number of 3D Gaussians. This improvement extends the applicability of 3D GS to handle intricate scenarios with specular and anisotropic surfaces.

Open-world instance-level scene understanding aims to locate and recognize unseen object categories that are not present in the annotated dataset. This task is challenging because the model needs to both localize novel 3D objects and infer their semantic categories. A key factor for the recent progress in 2D open-world perception is the availability of large-scale image-text pairs from the Internet, which cover a wide range of vocabulary concepts. However, this success is hard to replicate in 3D scenarios due to the scarcity of 3D-text pairs. To address this challenge, we propose to harness pre-trained vision-language (VL) foundation models that encode extensive knowledge from image-text pairs to generate captions for multi-view images of 3D scenes. This allows us to establish explicit associations between 3D shapes and semantic-rich captions. Moreover, to enhance the fine-grained visual-semantic representation learning from captions for object-level categorization, we design hierarchical point-caption association methods to learn semantic-aware embeddings that exploit the 3D geometry between 3D points and multi-view images. In addition, to tackle the localization challenge for novel classes in the open-world setting, we develop debiased instance localization, which involves training object grouping modules on unlabeled data using instance-level pseudo supervision. This significantly improves the generalization capabilities of instance grouping and, thus, the ability to accurately locate novel objects. We conduct extensive experiments on 3D semantic, instance, and panoptic segmentation tasks, covering indoor and outdoor scenes across three datasets. Our method outperforms baseline methods by a significant margin in semantic segmentation (e.g. 34.5% ∼ 65.3%), instance segmentation (e.g. 21.8% ∼ 54.0%), and panoptic segmentation (e.g. 14.7% ∼ 43.3%). Code will be available.

Tracking visual objects from a single initial exemplar in the testing phase has been broadly cast as a one-/few-shot problem, i.e., one-shot learning for initial adaptation and few-shot learning for online adaptation. The recent few-shot online adaptation methods incorporate the prior knowledge from large amounts of annotated training data via complex meta-learning optimization in the offline phase. This helps the online deep trackers to achieve fast adaptation and reduce overfitting risk in tracking. In this paper, we propose a simple yet effective recursive least-squares estimator-aided online learning approach for few-shot online adaptation without requiring offline training. It allows an in-built memory retention mechanism for the model to remember the knowledge about the object seen before, and thus the seen data can be safely removed from training. This also bears certain similarities to the emerging continual learning field in preventing catastrophic forgetting. This mechanism enables us to unveil the power of modern online deep trackers without incurring too much extra computational cost. We evaluate our approach based on two networks in the online learning families for tracking, i.e., multi-layer perceptrons in RT-MDNet and convolutional neural networks in DiMP. The consistent improvements on several challenging tracking benchmarks demonstrate its effectiveness and efficiency.

Depth-aware Video Panoptic Segmentation (DVPS) is a challenging task that requires predicting the semantic class and 3D depth of each pixel in a video, while also segmenting and consistently tracking objects across frames. Predominant methodologies treat this as a multi-task learning problem, tackling each constituent task independently, thus restricting their capacity to leverage interrelationships amongst tasks and requiring parameter tuning for each task. To surmount these constraints, we present Slot-IVPS, a new approach employing an object-centric model to acquire unified object representations, thereby facilitating the model's ability to simultaneously capture semantic and depth information. Specifically, we introduce a novel representation, Integrated Panoptic Slots (IPS), to capture both semantic and depth information for all panoptic objects within a video, encompassing background semantics and foreground instances. Subsequently, we propose an integrated feature generator and enhancer to extract depth-aware features, alongside the Integrated Video Panoptic Retriever (IVPR), which iteratively retrieves spatial-temporal coherent object features and encodes them into IPS. The resulting IPS can be effortlessly decoded into an array of video outputs, including depth maps, classifications, masks, and object instance IDs. We undertake comprehensive analyses across four datasets, attaining state-of-the-art performance in both Depth-aware Video Panoptic Segmentation and Video Panoptic Segmentation tasks.

The performance of existing sign language recognition approaches is typically limited by the scale of training data. To address this issue, we propose a mutual enhancement network (MEN) for joint sign language recognition and education. First, a sign language recognition system built upon a spatial-temporal network is proposed to recognize the semantic category of a given sign language video. Besides, a sign language education system is developed to detect the failure modes of learners and further guide them to sign correctly. Our theoretical contribution lies in formulating the above two systems as an estimation-maximization (EM) framework, which can progressively boost each other. The recognition system could become more robust and accurate with more training data collected by the education system, while the education system could guide the learners to sign more precisely, benefiting from the hand shape analysis module of the recognition system. Experimental results on three large-scale sign language recognition datasets validate the superiority of the proposed framework.

In this paper, we present an implicit surface reconstruction method with 3D Gaussian Splatting (3DGS), namely 3DGSR, that allows for accurate 3D reconstruction with intricate details while inheriting the high efficiency and rendering quality of 3DGS. The key insight is to incorporate an implicit signed distance field (SDF) within 3D Gaussians for surface modeling, and to enable the alignment and joint optimization of both SDF and 3D Gaussians. To achieve this, we design coupling strategies that align and associate the SDF with 3D Gaussians, allowing for unified optimization and enforcing surface constraints on the 3D Gaussians. With alignment, optimizing the 3D Gaussians provides supervisory signals for SDF learning, enabling the reconstruction of intricate details. However, this only offers sparse supervisory signals to the SDF at locations occupied by Gaussians, which is insufficient for learning a continuous SDF. Then, to address this limitation, we incorporate volumetric rendering and align the rendered geometric attributes (depth, normal) with that derived from 3DGS. In sum, these two designs allow SDF and 3DGS to be aligned, jointly optimized, and mutually boosted. Our extensive experimental results demonstrate that our 3DGSR enables high-quality 3D surface reconstruction while preserving the efficiency and rendering quality of 3DGS. Besides, our method competes favorably with leading surface reconstruction techniques while offering a more efficient learning process and much better rendering qualities.

While high-quality texture maps are essential for realistic 3D asset rendering, few studies have explored learning directly in the texture space, especially on large-scale datasets. In this work, we depart from the conventional approach of relying on pre-trained 2D diffusion models for testtime optimization of 3D textures. Instead, we focus on the fundamental problem of learning in the UV texture space itself. For the first time, we train a large diffusion model capable of directly generating high-resolution texture maps in a feed-forward manner. To facilitate efficient learning in high-resolution UV spaces, we propose a scalable network architecture that interleaves convolutions on UV maps with attention layers on point clouds. Leveraging this architectural design, we train a 700 million parameter diffusion model that can generate UV texture maps guided by text prompts and single-view images. Once trained, our model naturally supports various extended applications, including text-guided texture inpainting, sparse-view texture completion, and text-driven texture synthesis. The code is available at https://github.com/CVMI-Lab/TEXGen.

Weakly supervised detection of anomalies in surveillance videos is a challenging task. Going beyond existing works that have deficient capabilities to localize anomalies in long videos, we propose a novel glance and focus network to effectively integrate spatial-temporal information for accurate anomaly detection. In addition, we empirically found that existing approaches that use feature magnitudes to represent the degree of anomalies typically ignore the effects of scene variations, and hence result in sub-optimal performance due to the inconsistency of feature magnitudes across scenes. To address this issue, we propose the Feature Amplification Mechanism and a Magnitude Contrastive Loss to enhance the discriminativeness of feature magnitudes for detecting anomalies. Experimental results on two large-scale benchmarks UCF-Crime and XD-Violence manifest that our method outperforms state-of-the-art approaches.

3D automatic annotation has received increased attention since manually annotating 3D point clouds is laborious. However, existing methods are usually complicated, e.g., pipelined training for 3D foreground/background segmentation, cylindrical object proposals, and point completion. Furthermore, they often overlook the inter-object feature correlation that is particularly informative to hard samples for 3D annotation. To this end, we propose a simple yet effective end-to-end Context-Aware Transformer (CAT) as an automated 3D-box labeler to generate precise 3D box annotations from 2D boxes, trained with a small number of human annotations. We adopt the general encoder-decoder architecture, where the CAT encoder consists of an intra-object encoder (local) and an inter-object encoder (global), performing self-attention along the sequence and batch dimensions, respectively. The former models intra-object interactions among points and the latter extracts feature relations among different objects, thus boosting scene-level understanding. Via local and global encoders, CAT can generate high-quality 3D box annotations with a streamlined workflow, allowing it to outperform existing state-of-the-arts by up to 1.79% 3D AP on the hard task of the KITTI test set.

Semantic segmentation is still a challenging task for parsing diverse contexts in different scenes, thus the fixed classifier might not be able to well address varying feature distributions during testing. Different from the mainstream literature where the efficacy of strong backbones and effective decoder heads has been well studied, in this paper, additional contextual hints are instead exploited via learning a context-aware classifier whose content is data-conditioned, decently adapting to different latent distributions. Since only the classifier is dynamically altered, our method is model-agnostic and can be easily applied to generic segmentation models. Notably, with only negligible additional parameters and +2\% inference time, decent performance gain has been achieved on both small and large models with challenging benchmarks, manifesting substantial practical merits brought by our simple yet effective method. The implementation is available at https://github.com/tianzhuotao/CAC.

By eliminating compute operations intelligently based on the run time input, dynamic pruning (DP) promises to improve deep neural network inference speed substantially without incurring a major impact on their accuracy. Although many DP algorithms with good pruning performance have been proposed, it remains a challenge to translate these theoretical reductions in compute operations into satisfactory end-to-end speedups in practical real-world implementations. The overhead of identifying operations to be pruned during run time, the need to efficiently process the resulting dynamic dataflow, and the non-trivial memory I/O bottleneck that emerges as the number of compute operations reduces, have all contributed to the challenge of implementing practical DP systems.

Recent advance in 2D CNNs has revealed that large kernels are important. However, when directly applying large convolutional kernels in 3D CNNs, severe difficulties are met, where those successful module designs in 2D become surprisingly ineffective on 3D networks, including the popular depth-wise convolution. To address this vital challenge, we instead propose the spatial-wise partition convolution and its large-kernel module. As a result, it avoids the optimization and efficiency issues of naive 3D large kernels. Our large-kernel 3D CNN network, LargeKernel3D, yields notable improvement in 3D tasks of semantic segmentation and object detection. It achieves 73.9% mIoU on the ScanNetv2 semantic segmentation and 72.8% NDS nuScenes object detection benchmarks, ranking 1st on the nuScenes LIDAR leaderboard. The performance further boosts to 74.2% NDS with a simple multi-modal fusion. In addition, LargeKernel3D can be scaled to 17×17×17 kernel size on Waymo 3D object detection. For the first time, we show that large kernels are feasible and essential for 3D visual tasks. Our code and models is available at github.com/dvlab-research/LargeKernel3D.

3D object detectors usually rely on hand-crafted proxies, e.g., anchors or centers, and translate well-studied 2D frameworks to 3D. Thus, sparse voxel features need to be densified and processed by dense prediction heads, which inevitably costs extra computation. In this paper, we instead propose VoxelNext for fully sparse 3D object detection. Our core insight is to predict objects directly based on sparse voxel features, without relying on hand-crafted proxies. Our strong sparse convolutional network VoxelNeXt detects and tracks 3D objects through voxel features entirely. It is an elegant and efficient framework, with no need for sparse-to-dense conversion or NMS post-processing. Our method achieves a better speed-accuracy trade-off than other mainframe detectors on the nuScenes dataset. For the first time, we show that a fully sparse voxel-based representation works decently for LIDAR 3D object detection and tracking. Extensive experiments on nuScenes, Waymo, and Argoverse2 benchmarks validate the effectiveness of our approach. Without bells and whistles, our model outperforms all existing LIDAR methods on the nuScenes tracking test benchmark. Code and models are available at github.com/dvlab-research/VoxelNeXt.

We present Cart, a new approach towards articulatedobject manipulations by human commands. Beyond the existing work that focuses on inferring articulation structures, we further support manipulating articulated shapes to align them subject to simple command templates. The key of Cart is to utilize the prediction of object structures to connect visual observations with user commands for effective manipulations. It is achieved by encoding command messages for motion prediction and a test-time adaptation to adjust the amount of movement from only command supervision. For a rich variety of object categories, Cart can accurately manipulate object shapes and outperform the state-of-the-art approaches in understanding the inherent articulation structures. Also, it can well generalize to unseen object categories and real-world objects. We hope Cart could open new directions for instructing machines to operate articulated objects. Code is available at https://github.com/dvlab-research/Cart.

Rendering novel view images is highly desirable for many applications. Despite recent progress, it remains challenging to render high-fidelity and view-consistent novel views of large-scale scenes from in-the-wild images with inevitable artifacts (e.g., motion blur). To this end, we develop a hybrid neural rendering model that makes image-based representation and neural 3D representation join forces to render high-quality, view-consistent images. Besides, images captured in the wild inevitably contain artifacts, such as motion blur, which deteriorates the quality of rendered images. Accordingly, we propose strategies to simulate blur effects on the rendered images to mitigate the negative influence of blurriness images and reduce their importance during training based on precomputed quality-aware weights. Extensive experiments on real and synthetic data demonstrate our model surpasses state-of-the-art point-based methods for novel view synthesis. The code is available at https://daipengwa.github.io/Hybrid-Rendering-ProjectPage/.

Open-vocabulary scene understanding aims to localize and recognize unseen categories beyond the annotated label space. The recent breakthrough of 2D open-vocabulary perception is largely driven by Internet-scale paired image-text data with rich vocabulary concepts. However, this success cannot be directly transferred to 3D scenarios due to the inaccessibility of large-scale 3D-text pairs. To this end, we propose to distill knowledge encoded in pretrained vision-language (VL) foundation models through captioning multi-view images from 3D, which allows explicitly associating 3D and semantic-rich captions. Further, to foster coarse-to-fine visual-semantic representation learning from captions, we design hierarchical 3D-caption pairs, leveraging geometric constraints between 3D scenes and multi-view images. Finally, by employing contrastive learning, the model learns language-aware embeddings that connect 3D and text for open-vocabulary tasks. Our method not only remarkably outperforms baseline methods by 25.8% ~ 44.7% hIoU and 14.5% ~ 50.4% hAP50in open-vocabulary semantic and instance segmentation, but also shows robust transferability on challenging zero-shot domain transfer tasks. See the project website at https://dingry.github.io/projects/PLA.

3D semantic segmentation on multi-scan large-scale point clouds plays an important role in autonomous systems. Unlike the single-scan-based semantic segmentation task, this task requires distinguishing the motion states of points in addition to their semantic categories. However, methods designed for single-scan-based segmentation tasks perform poorly on the multi-scan task due to the lacking of an effective way to integrate temporal information. We propose MarS3D, a plug-and-play motion-aware module for semantic segmentation on multi-scan 3D point clouds. This module can be flexibly combined with single-scan models to allow them to have multi-scan perception abilities. The model encompasses two key designs: the Cross-Frame Feature Embedding module for enriching representation learning and the Motion-Aware Feature Learning module for enhancing motion awareness. Extensive experiments show that MarS3D can improve the performance of the baseline model by a large margin. The code is available at https://github.com/CVMI-Lab/MarS3D.

A recent study has shown a phenomenon called neural collapse in that the within-class means of features and the classifier weight vectors converge to the vertices of a simplex equiangular tight frame at the terminal phase of training for classification. In this paper, we explore the cor-responding structures of the last-layer feature centers and classifiers in semantic segmentation. Based on our empirical and theoretical analysis, we point out that semantic segmentation naturally brings contextual correlation and imbalanced distribution among classes, which breaks the equiangular and maximally separated structure of neural collapse for both feature centers and classifiers. However, such a symmetric structure is beneficial to discrimination for the minor classes. To preserve these advantages, we in-troduce a regularizer on feature centers to encourage the network to learn features closer to the appealing structure in imbalanced semantic segmentation. Experimental results show that our method can bring significant improvements on both 2D and 3D semantic segmentation bench-marks. Moreover, our method ranks 1stand sets a new record (+6.8% mIoU) on the ScanNet200 test leaderboard.

In this work, we focus on synthesizing high-quality textures on 3D meshes. We present Point-UV diffusion, a coarse-to-fine pipeline that marries the denoising diffusion model with UV mapping to generate 3D consistent and high-quality texture images in UV space. We start with introducing a point diffusion model to synthesize low-frequency texture components with our tailored style guidance to tackle the biased color distribution. The derived coarse texture offers global consistency and serves as a condition for the subsequent UV diffusion stage, aiding in regularizing the model to generate a 3D consistent UV texture image. Then, a UV diffusion model with hybrid conditions is developed to enhance the texture fidelity in the 2D UV space. Our method can process meshes of any genus, generating diversified, geometry-compatible, and high-fidelity textures. Code is available at https://cvmi-lab.github.io/Point-UV-Diffusion.

Category-level 6D pose estimation aims to predict the poses and sizes of unseen objects from a specific category. Thanks to prior deformation, which explicitly adapts a category-specific 3D prior (i.e., a 3D template) to a given object instance, prior-based methods attained great success and have become a major research stream. However, obtaining category-specific priors requires collecting a large amount of 3D models, which is labor-consuming and often not accessible in practice. This motivates us to investigate whether priors are necessary to make prior-based methods effective. Our empirical study shows that the 3D prior itself is not the credit to the high performance. The keypoint actually is the explicit deformation process, which aligns camera and world coordinates supervised by world-space 3D models (also called canonical space). Inspired by these observations, we introduce a simple prior-free implicit space transformation network, namely IST-Net, to transform camera-space features to world-space counterparts and build correspondences between them in an implicit manner without relying on 3D priors. Besides, we design camera- and world-space enhancers to enrich the features with pose-sensitive information and geometrical constraints, respectively. Albeit simple, IST-Net achieves state-of-the-art performance based-on prior-free design, with top inference speed on the REAL275 benchmark. Our code and models are available at https://github.com/CVMI-Lab/IST-Net.

Implicit neural rendering, using signed distance function (SDF) representation with geometric priors like depth or surface normal, has made impressive strides in the surface reconstruction of large-scale scenes. However, applying this method to reconstruct a room-level scene from images may miss structures in low-intensity areas and/or small, thin objects. We have conducted experiments on three datasets to identify limitations of the original color rendering loss and priors-embedded SDF scene representation.Our findings show that the color rendering loss creates an optimization bias against low-intensity areas, resulting in gradient vanishing and leaving these areas unoptimized. To address this issue, we propose a feature-based color rendering loss that utilizes non-zero feature values to bring back optimization signals. Additionally, the SDF representation can be influenced by objects along a ray path, disrupting the monotonic change of SDF values when a single object is present. Accordingly, we explore using the occupancy representation, which encodes each point separately and is unaffected by objects along a querying ray. Our experimental results demonstrate that the joint forces of the feature-based rendering loss and Occ-SDF hybrid representation scheme can provide high-quality reconstruction results, especially in challenging room-level scenarios. The code is available at https://github.com/shawLyu/Occ-SDF-Hybrid

Generalized Category Discovery (GCD) aims to discover novel categories in unlabelled datasets using knowledge learned from labelled samples. Previous studies argued that parametric classifiers are prone to overfitting to seen categories, and endorsed using a non-parametric classifier formed with semi-supervised k-means. However, in this study, we investigate the failure of parametric classifiers, verify the effectiveness of previous design choices when high-quality supervision is available, and identify unreliable pseudo-labels as a key problem. We demonstrate that two prediction biases exist: the classifier tends to predict seen classes more often, and produces an imbalanced distribution across seen and novel categories. Based on these findings, we propose a simple yet effective parametric classification method that benefits from entropy regularisation, achieves state-of-the-art performance on multiple GCD benchmarks and shows strong robustness to unknown class numbers. We hope the investigation and proposed simple framework can serve as a strong baseline to facilitate future studies in this field. Our code is available at: https://github.com/CVMI-Lab/SimGCD.

Synthesizing realistic videos according to a given speech is still an open challenge. Previous works have been plagued by issues such as inaccurate lip shape generation and poor image quality. The key reason is that only motions and appearances on limited facial areas (e.g., lip area) are mainly driven by the input speech. Therefore, directly learning a mapping function from speech to the entire head image is prone to ambiguity, particularly when using a short video for training. We thus propose a decomposition-synthesis-composition framework named Speech to Lip (Speech2Lip) that disentangles speech-sensitive and speech-insensitive motion/appearance to facilitate effective learning from limited training data, resulting in the generation of natural-looking videos. First, given a fixed head pose (i.e., canonical space), we present a speech-driven implicit model for lip image generation which concentrates on learning speech-sensitive motion and appearance. Next, to model the major speech-insensitive motion (i.e., head movement), we introduce a geometry-aware mutual explicit mapping (GAMEM) module that establishes geometric mappings between different head poses. This allows us to paste generated lip images at the canonical space onto head images with arbitrary poses and synthesize talking videos with natural head movements. In addition, a Blend-Net and a contrastive sync loss are introduced to enhance the overall synthesis performance. Quantitative and qualitative results on three benchmarks demonstrate that our model can be trained by a video of just a few minutes in length and achieve state-of-the-art performance in both visual quality and speechvisual synchronization. Code: https://github.com/CVMILab/Speech2Lip.

Deriving reliable region-word alignment from image-text pairs is critical to learn object-level vision-language representations for open-vocabulary object detection. Existing methods typically rely on pre-trained or self-trained vision-language models for alignment, which are prone to limitations in localization accuracy or generalization capabilities. In this paper, we propose CoDet, a novel approach that overcomes the reliance on pre-aligned vision-language space by reformulating region-word alignment as a co-occurring object discovery problem. Intuitively, by grouping images that mention a shared concept in their captions, objects corresponding to the shared concept shall exhibit high co-occurrence among the group. CoDet then leverages visual similarities to discover the co-occurring objects and align them with the shared concept. Extensive experiments demonstrate that CoDet has superior performances and compelling scalability in open-vocabulary detection, e.g., by scaling up the visual backbone, CoDet achieves 37.0 $AP^m_{novel}$ and 44.7 $AP^m_{all}$ on OV-LVIS, surpassing the previous SoTA by 4.2 $AP^m_{novel}$ and 9.8 $AP^m_{all}$. Code is available at https://github.com/CVMI-Lab/CoDet.

In this paper, we propose a novel data-pruning approach called moving-one-sample-out (MoSo), which aims to identify and remove the least informative samples from the training set. The core insight behind MoSo is to determine the importance of each sample by assessing its impact on the optimal empirical risk. This is achieved by measuring the extent to which the empirical risk changes when a particular sample is excluded from the training set. Instead of using the computationally expensive leaving-one-out-retraining procedure, we propose an efficient first-order approximator that only requires gradient information from different training stages. The key idea behind our approximation is that samples with gradients that are consistently aligned with the average gradient of the training set are more informative and should receive higher scores, which could be intuitively understood as follows: if the gradient from a specific sample is consistent with the average gradient vector, it implies that optimizing the network using the sample will yield a similar effect on all remaining samples. Experimental results demonstrate that MoSo effectively mitigates severe performance degradation at high pruning ratios and achieves satisfactory performance across various settings. Experimental results demonstrate that MoSo effectively mitigates severe performance degradation at high pruning ratios and outperforms state-of-the-art methods by a large margin across various settings.

Existing methods for adapting Neural Radiance Fields (NeRFs) to scene changes require extensive data capture and model retraining, which is both time-consuming and labor-intensive. In this paper, we tackle the challenge of efficiently adapting NeRFs to real-world scene changes over time using a few new images while retaining the memory of unaltered areas, focusing on the continual learning aspect of NeRFs. To this end, we propose CL-NeRF, which consists of two key components: a lightweight expert adaptor for adapting to new changes and evolving scene representations and a conflict-aware knowledge distillation learning objective for memorizing unchanged parts. We also present a new benchmark for evaluating Continual Learning of NeRFs with comprehensive metrics. Our extensive experiments demonstrate that CL-NeRF can synthesize high-quality novel views of both changed and unchanged regions with high training efficiency, surpassing existing methods in terms of reducing forgetting and adapting to changes. Code and benchmark will be made available.

In this work, we report the set-up and results of the Liver Tumor Segmentation Benchmark (LiTS), which was organized in conjunction with the IEEE International Symposium on Biomedical Imaging (ISBI) 2017 and the International Conferences on Medical Image Computing and Computer-Assisted Intervention (MICCAI) 2017 and 2018. The image dataset is diverse and contains primary and secondary tumors with varied sizes and appearances with various lesion-to-background levels (hyper-/hypo-dense), created in collaboration with seven hospitals and research institutions. Seventy-five submitted liver and liver tumor segmentation algorithms were trained on a set of 131 computed tomography (CT) volumes and were tested on 70 unseen test images acquired from different patients. We found that not a single algorithm performed best for both liver and liver tumors in the three events. The best liver segmentation algorithm achieved a Dice score of 0.963, whereas, for tumor segmentation, the best algorithms achieved Dices scores of 0.674 (ISBI 2017), 0.702 (MICCAI 2017), and 0.739 (MICCAI 2018). Retrospectively, we performed additional analysis on liver tumor detection and revealed that not all top-performing segmentation algorithms worked well for tumor detection. The best liver tumor detection method achieved a lesion-wise recall of 0.458 (ISBI 2017), 0.515 (MICCAI 2017), and 0.554 (MICCAI 2018), indicating the need for further research. LiTS remains an active benchmark and resource for research, e.g., contributing the liver-related segmentation tasks in http://medicaldecathlon.com/. In addition, both data and online evaluation are accessible via https://competitions.codalab.org/competitions/17094.

In this paper, we present a conceptually simple, strong, and efficient framework for fully- and weakly-supervised panoptic segmentation, called Panoptic FCN. Our approach aims to represent and predict foreground things and background stuff in a unified fully convolutional pipeline, which can be optimized with point-based fully or weak supervision. In particular, Panoptic FCN encodes each object instance or stuff category with the proposed kernel generator and produces the prediction by convolving the high-resolution feature directly. With this approach, instance-aware and semantically consistent properties for things and stuff can be respectively satisfied in a simple generate-kernel-then-segment workflow. Without extra boxes for localization or instance separation, the proposed approach outperforms the previous box-based and -free models with high efficiency. Furthermore, we propose a new form of point-based annotation for weakly-supervised panoptic segmentation. It only needs several random points for both things and stuff, which dramatically reduces the annotation cost of human. The proposed Panoptic FCN is also proved to have much superior performance in this weakly-supervised setting, which achieves 82% of the fully-supervised performance with only 20 randomly annotated points per instance. Extensive experiments demonstrate the effectiveness and efficiency of Panoptic FCN on COCO, VOC 2012, Cityscapes, and Mapillary Vistas datasets. And it sets up a new leading benchmark for both fully- and weakly-supervised panoptic segmentation.

This paper presents a new text-guided 3D shape generation approach DreamStone that uses images as a stepping stone to bridge the gap between the text and shape modalities for generating 3D shapes without requiring paired text and 3D data. The core of our approach is a two-stage feature-space alignment strategy that leverages a pre-trained single-view reconstruction (SVR) model to map CLIP features to shapes: to begin with, map the CLIP image feature to the detail-rich 3D shape space of the SVR model, then map the CLIP text feature to the 3D shape space through encouraging the CLIP-consistency between the rendered images and the input text. Besides, to extend beyond the generative capability of the SVR model, we design the text-guided 3D shape stylization module that can enhance the output shapes with novel structures and textures. Further, we exploit pre-trained text-to-image diffusion models to enhance the generative diversity, fidelity, and stylization capability. Our approach is generic, flexible, and scalable. It can be easily integrated with various SVR models to expand the generative space and improve the generative fidelity. Extensive experimental results demonstrate that our approach outperforms the state-of-the-art methods in terms of generative quality and consistency with the input text.

Although considerable progress has been obtained in neural network quantization for efficient inference, existing methods are not scalable to heterogeneous devices as one dedicated model needs to be trained, transmitted, and stored for one specific hardware setting, incurring considerable costs in model training and maintenance. In this paper, we study a new vertical-layered representation of neural network weights for encapsulating all quantized models into a single one. It represents weights as a group of bits (i.e., vertical layers) organized from the most significant bit (also called the basic layer) to less significant bits (i.e., enhance layers). Hence, a neural network with an arbitrary quantization precision can be obtained by adding corresponding enhance layers to the basic layer. However, we empirically find that models obtained with existing quantization methods suffer severe performance degradation if they are adapted to vertical-layered weight representation. To this end, we propose a simple once quantization-aware training (QAT) scheme for obtaining high-performance vertical-layered models. Our design incorporates a cascade downsampling mechanism with the multi-objective optimization employed to train the shared source model weights such that they can be updated simultaneously, considering the performance of all networks. After the model is trained, to construct a vertical-layered network, the lowest bit-width quantized weights become the basic layer, and every bit dropped along the downsampling process act as an enhance layer. Our design is extensively evaluated on CIFAR-100 and ImageNet datasets. Experiments show that the proposed vertical-layered representation and developed once QAT scheme are effective in embodying multiple quantized networks into a single one and allow one-time training, and it delivers comparable performance as that of quantized models tailored to any specific bit-width.

In this paper, we present a self-training method, named ST3D++, with a holistic pseudo label denoising pipeline for unsupervised domain adaptation on 3D object detection. ST3D++ aims at reducing noise in pseudo label generation as well as alleviating the negative impacts of noisy pseudo labels on model training. First, ST3D++ pre-trains the 3D object detector on the labeled source domain with random object scaling (ROS) which is designed to reduce target domain pseudo label noise arising from object scale bias of the source domain. Then, the detector is progressively improved through alternating between generating pseudo labels and training the object detector with pseudo-labeled target domain data. Here, we equip the pseudo label generation process with a hybrid quality-aware triplet memory to improve the quality and stability of generated pseudo labels. Meanwhile, in the model training stage, we propose a source data assisted training strategy and a curriculum data augmentation policy to effectively rectify noisy gradient directions and avoid model over-fitting to noisy pseudo labeled data. These specific designs enable the detector to be trained on meticulously refined pseudo labeled target data with denoised training signals, and thus effectively facilitate adapting an object detector to a target domain without requiring annotations. Finally, our method is assessed on four 3D benchmark datasets (i.e., Waymo, KITTI, Lyft, and nuScenes) for three common categories (i.e., car, pedestrian and bicycle). ST3D++ achieves state-of-the-art performance on all evaluated settings, outperforming the corresponding baseline by a large margin (e.g., 9.6% ∼ 38.16% on Waymo → KITTI in terms of AP[Formula: see text]), and even surpasses the fully supervised oracle results on the KITTI 3D object detection benchmark with target prior. Code is available at https://github.com/CVMI-Lab/ST3D.

This paper presents a new text-guided technique for generating 3D shapes. The technique leverages a hybrid 3D shape representation, namely EXIM, combining the strengths of explicit and implicit representations. Specifically, the explicit stage controls the topology of the generated 3D shapes and enables local modifications, whereas the implicit stage refines the shape and paints it with plausible colors. Also, the hybrid approach separates the shape and color and generates color conditioned on shape to ensure shape-color consistency. Unlike the existing state-of-the-art methods, we achieve high-fidelity shape generation from natural-language descriptions without the need for time-consuming per-shape optimization or reliance on human-annotated texts during training or test-time optimization. Further, we demonstrate the applicability of our approach to generate indoor scenes with consistent styles using text-induced 3D shapes. Through extensive experiments, we demonstrate the compelling quality of our results and the high coherency of our generated shapes with the input texts, surpassing the performance of existing methods by a significant margin. Codes and models are released at https://github.com/liuzhengzhe/EXIM.

We explore the way to alleviate the label-hungry problem in a semi-supervised setting for 3D instance segmentation. To leverage the unlabeled data to boost model performance, we present a novel Two-Way Inter-label Self-Training framework named TWIST. It exploits inherent correlations between semantic understanding and instance information of a scene. Specifically, we consider two kinds of pseudo labels for semantic- and instance-level supervision. Our key design is to provide object-level information for denoising pseudo labels and make use of their correlation for two-way mutual enhancement, thereby iteratively promoting the pseudo-label qualities. TWIST attains leading performance on both ScanNet and S3DIS, compared to recent 3D pre-training approaches, and can cooperate with them to further enhance performance, e.g., +4.4% AP50on 1%-label ScanNet data-efficient benchmark. Code is available at https://github.com/dvlab-research/TWIST.

Moiré patterns, appearing as color distortions, severely degrade image and video qualities when filming a screen with digital cameras. Considering the increasing demands for capturing videos, we study how to remove such undesirable moiré patterns in videos, namely video demoiréing. To this end, we introduce the first hand-held video demoiréing dataset with a dedicated data collection pipeline to ensure spatial and temporal alignments of captured data. Further, a baseline video demoiréing model with implicit feature space alignment and selective feature aggregation is developed to leverage complementary information from nearby frames to improve frame-level video demoiréing. More importantly, we propose a relation-based temporal consistency loss to encourage the model to learn temporal consistency priors directly from ground-truth reference videos, which facilitates producing temporally consistent predictions and effectively maintains frame-level qualities. Extensive experiments manifest the superiority of our model. Code is available at ht tps:// daipengwa. github.io/VDmoire_ProjectPage/.

Large-scale pre-training has been proven to be crucial for various computer vision tasks. However, with the increase of pre-training data amount, model architecture amount, and the private/inaccessible data, it is not very efficient or possible to pre-train all the model architectures on large-scale datasets. In this work, we investigate an alternative strategy for pre-training, namely Knowledge Distillation as Efficient Pre-training (KDEP), aiming to efficiently transfer the learned feature representation from existing pre-trained models to new student models for future downstream tasks. We observe that existing Knowledge Distillation (KD) methods are unsuitable towards pre-training since they normally distill the logits that are going to be discarded when transferred to downstream tasks. To resolve this problem, we propose a feature-based KD method with non-parametric feature dimension aligning. Notably, our method performs comparably with supervised pre-training counterparts in 3 downstream tasks and 9 downstream datasets requiring 10× less data and 5× less pre-training time. Code is available at https://github.com/CVMI-Lab/KDEP.

3D point cloud segmentation has made tremendous progress in recent years. Most current methods focus on aggregating local features, but fail to directly model long-range dependencies. In this paper, we propose Stratified Transformer that is able to capture long-range contexts and demonstrates strong generalization ability and high performance. Specifically, we first put forward a novel key sampling strategy. For each query point, we sample nearby points densely and distant points sparsely as its keys in a stratified way, which enables the model to enlarge the effective receptive field and enjoy long-range contexts at a low computational cost. Also, to combat the challenges posed by irregular point arrangements, we propose first-layer point embedding to aggregate local information, which facilitates convergence and boosts performance. Besides, we adopt contextual relative position encoding to adaptively capture position information. Finally, a memory-efficient implementation is introduced to overcome the issue of varying point numbers in each window. Extensive experiments demonstrate the effectiveness and superiority of our method on S3DIS, ScanNetv2 and ShapeNetPart datasets. Code is available at https://github.com/dvlab-research/Stratified-Transformer.

In this work, we present a conceptually simple yet effective framework for cross-modality 3D object detection, named voxel field fusion. The proposed approach aims to maintain cross-modality consistency by representing and fusing augmented image features as a ray in the voxel field. To this end, the learnable sampler is first designed to sample vital features from the image plane that are projected to the voxel grid in a point-to-ray manner, which maintains the consistency in feature representation with spatial context. In addition, ray-wise fusion is conducted to fuse features with the supplemental context in the constructed voxel field. We further develop mixed augmentor to align feature-variant transformations, which bridges the modality gap in data augmentation. The proposed framework is demonstrated to achieve consistent gains in various bench-marks and outperforms previous fusion-based methods on KITTI and nuScenes datasets. Code is made available at https://github.com/dvlab-research/VFF11Part of the work was done in MEGVII Research..

In this work, we explore the challenging task of generating 3D shapes from text. Beyond the existing works, we propose a new approach for text-guided 3D shape generation, capable of producing high-fidelity shapes with colors that match the given text description. This work has several technical contributions. First, we decouple the shape and color predictions for learning features in both texts and shapes, and propose the word-level spatial transformer to correlate word features from text with spatial features from shape. Also, we design a cyclic loss to encourage consistency between text and shape, and introduce the shape IMLE to diversify the generated shapes. Further, we extend the framework to enable text-guided shape manipulation. Extensive experiments on the largest existing text-shape benchmark [10] manifest the superiority of this work. The code and the models are available at https://github.com/liuzhengzhe/Towards-Implicit-Text-Guided-Shape-Generation.

To interpret deep networks, one main approach is to associate neurons with human-understandable concepts. However, existing methods often ignore the inherent connections of different concepts (e.g., dog and cat both belong to animals), and thus lose the chance to explain neurons responsible for higher-level concepts (e.g., animal). In this paper, we study hierarchical concepts inspired by the hierarchical cognition process of human beings. To this end, we propose HIerarchical Neuron concepT explainer (HINT) to effectively build bidirectional associations between neurons and hierarchical concepts in a low-cost and scalable manner. HINT enables us to systematically and quantitatively study whether and how the implicit hierarchical relationships of concepts are embedded into neurons. Specifically, HINT identifies collaborative neurons responsible for one concept and multimodal neurons pertinent to different concepts, at different semantic levels from concrete concepts (e.g., dog) to more abstract ones (e.g., animal). Finally, we verify the faithfulness of the associations using Weakly Supervised Object Localization, and demonstrate its applicability in various tasks, such as discovering saliency regions and explaining adversarial attacks. Code is available on https://github.com/AntonotnaWang/HINT.

In this paper, we propose a new query-based detection framework for crowd detection. Previous query-based detectors suffer from two drawbacks: first, multiple predictions will be inferred for a single object, typically in crowded scenes; second, the performance saturates as the depth of the decoding stage increases. Benefiting from the nature of the one-to-one label assignment rule, we propose a progressive predicting method to address the above issues. Specifically, we first select accepted queries prone to generate true positive predictions, then refine the rest noisy queries according to the previously accepted predictions. Experiments show that our method can significantly boost the performance of query-based detectors in crowded scenes. Equipped with our approach, Sparse RCNN achieves 92.0% AP, 41.4% MR−2and 83.2% JI on the challenging CrowdHuman [35] dataset, outperforming the box-based method MIP [8] that specifies in handling crowded scenarios. Moreover, the proposed method, robust to crowdedness, can still obtain consistent improvements on moderately and slightly crowded datasets like CityPersons [47] and COCO [26]. Code will be made publicly available at https://github.com/megvii-model/Iter-E2EDET.

Video Panoptic Segmentation (VPS) aims at assigning a class label to each pixel, uniquely segmenting and identifying all object instances consistently across all frames. Classic solutions usually decompose the VPS task into several subtasks and utilize multiple surrogates (e.g. boxes and masks, centers and offsets) to represent objects. However, this divide-and-conquer strategy requires complex post-processing in both spatial and temporal domains and is vulnerable to failures from surrogate tasks. In this paper, inspired by object-centric learning which learns compact and robust object representations, we present Slot- VPS, the first end-to-end framework for this task. We encode all panoptic entities in a video, including both foreground instances and background semantics, with a unified representation called panoptic slots. The coherent spatio-temporal object's information is retrieved and encoded into the panoptic slots by the proposed Video Panoptic Retriever, enabling to localize, segment, differentiate, and associate objects in a unified manner. Finally, the output panoptic slots can be directly converted into the class, mask, and object ID of panoptic objects in the video. We conduct extensive ablation studies and demonstrate the effectiveness of our approach on two benchmark datasets, Cityscapes- VP S (val and test sets) and VIPER (val set), achieving new state-of-the-art performance of 63.7, 63.3 and 56.2 VPQ, respectively.

Manually annotating 3D point clouds is laborious and costly, limiting the training data preparation for deep learning in real-world object detection. While a few previous studies tried to automatically generate 3D bounding boxes from weak labels such as 2D boxes, the quality is sub-optimal compared to human annotators. This work proposes a novel autolabeler, called multimodal attention point generator (MAP-Gen), that generates high-quality 3D labels from weak 2D boxes. It leverages dense image information to tackle the sparsity issue of 3D point clouds, thus improving label quality. For each 2D pixel, MAP-Gen predicts its corresponding 3D coordinates by referencing context points based on their 2D semantic or geometric relationships. The generated 3D points densify the original sparse point clouds, followed by an encoder to regress 3D bounding boxes. Using MAP-Gen, object detection networks that are weakly supervised by 2D boxes can achieve 94 ~ 99% performance of those fully supervised by 3D annotations. It is hopeful this newly proposed MAP-Gen autolabeling flow can shed new light on utilizing multimodal information for enriching sparse point clouds.

In this work, we present a unified framework for multi-modality 3D object detection, named UVTR. The proposed method aims to unify multi-modality representations in the voxel space for accurate and robust single- or cross-modality 3D detection. To this end, the modality-specific space is first designed to represent different inputs in the voxel feature space. Different from previous work, our approach preserves the voxel space without height compression to alleviate semantic ambiguity and enable spatial connections. To make full use of the inputs from different sensors, the cross-modality interaction is then proposed, including knowledge transfer and modality fusion. In this way, geometry-aware expressions in point clouds and context-rich features in images are well utilized for better performance and robustness. The transformer decoder is applied to efficiently sample features from the unified space with learnable positions, which facilitates object-level interactions. In general, UVTR presents an early attempt to represent different modalities in a unified framework. It surpasses previous work in single- or multi-modality entries. The proposed method achieves leading performance in the nuScenes test set for both object detection and the following object tracking task. Code is made publicly available at https://github.com/dvlab-research/UVTR.

3D scenes are dominated by a large number of background points, which is redundant for the detection task that mainly needs to focus on foreground objects. In this paper, we analyze major components of existing sparse 3D CNNs and find that 3D CNNs ignores the redundancy of data and further amplifies it in the down-sampling process, which brings a huge amount of extra and unnecessary computational overhead. Inspired by this, we propose a new convolution operator named spatial pruned sparse convolution (SPS-Conv), which includes two variants, spatial pruned submanifold sparse convolution (SPSS-Conv) and spatial pruned regular sparse convolution (SPRS-Conv), both of which are based on the idea of dynamically determine crucial areas for performing computations to reduce redundancy. We empirically find that magnitude of features can serve as an important cues to determine crucial areas which get rid of the heavy computations of learning-based methods. The proposed modules can easily be incorporated into existing sparse 3D CNNs without extra architectural modifications. Extensive experiments on the KITTI and nuScenes datasets demonstrate that our method can achieve more than 50% reduction in GFLOPs without compromising the performance.

In this paper, we empirically study how to make the most of low-resolution frames for efficient video recognition. Existing methods mainly focus on developing compact networks or alleviating temporal redundancy of video inputs to increase efficiency, whereas compressing frame resolution has rarely been considered a promising solution. A major concern is the poor recognition accuracy on low-resolution frames. We thus start by analyzing the underlying causes of performance degradation on low-resolution frames. Our key finding is that the major cause of degradation is not information loss in the down-sampling process, but rather the mismatch between network architecture and input scale. Motivated by the success of knowledge distillation (KD), we propose to bridge the gap between network and input size via cross-resolution KD (ResKD). Our work shows that ResKD is a simple but effective method to boost recognition accuracy on low-resolution frames. Without bells and whistles, ResKD considerably surpasses all competitive methods in terms of efficiency and accuracy on four large-scale benchmark datasets, i.e., ActivityNet, FCVID, Mini-Kinetics, Something-Something V2. In addition, we extensively demonstrate its effectiveness over state-of-the-art architectures, i.e., 3D-CNNs and Video Transformers, and scalability towards super low-resolution frames. The results suggest ResKD can serve as a general inference acceleration method for state-of-the-art video recognition. Our code will be available at https://github.com/CVMI-Lab/ResKD.

In this paper, we tackle the problem of learning visual representations from unlabeled scene-centric data. Existing works have demonstrated the potential of utilizing the underlying complex structure within scene-centric data; still, they commonly rely on hand-crafted objectness priors or specialized pretext tasks to build a learning framework, which may harm generalizability. Instead, we propose contrastive learning from data-driven semantic slots, namely SlotCon, for joint semantic grouping and representation learning. The semantic grouping is performed by assigning pixels to a set of learnable prototypes, which can adapt to each sample by attentive pooling over the feature and form new slots. Based on the learned data-dependent slots, a contrastive objective is employed for representation learning, which enhances the discriminability of features, and conversely facilitates grouping semantically coherent pixels together. Compared with previous efforts, by simultaneously optimizing the two coupled objectives of semantic grouping and contrastive learning, our approach bypasses the disadvantages of hand-crafted priors and is able to learn object/group-level representations from scene-centric images. Experiments show our approach effectively decomposes complex scenes into semantic groups for feature learning and significantly benefits downstream tasks, including object detection, instance segmentation, and semantic segmentation. Code is available at: https://github.com/CVMI-Lab/SlotCon.

Despite substantial progress in 3D object detection, advanced 3D detectors often suffer from heavy computation overheads. To this end, we explore the potential of knowledge distillation (KD) for developing efficient 3D object detectors, focusing on popular pillar- and voxel-based detectors. In the absence of well-developed teacher-student pairs, we first study how to obtain student models with good trade offs between accuracy and efficiency from the perspectives of model compression and input resolution reduction. Then, we build a benchmark to assess existing KD methods developed in the 2D domain for 3D object detection upon six well-constructed teacher-student pairs. Further, we propose an improved KD pipeline incorporating an enhanced logit KD method that performs KD on only a few pivotal positions determined by teacher classification response and a teacher-guided student model initialization to facilitate transferring teacher model's feature extraction ability to students through weight inheritance. Finally, we conduct extensive experiments on the Waymo dataset. Our best performing model achieves $65.75\%$ LEVEL 2 mAPH surpassing its teacher model and requiring only $44\%$ of teacher flops. Our most efficient model runs 51 FPS on an NVIDIA A100, which is $2.2\times$ faster than PointPillar with even higher accuracy. Code will be available.

Most existing 3D point cloud object detection approaches heavily rely on large amounts of labeled training data. However, the labeling process is costly and time-consuming. This paper considers few-shot 3D point cloud object detection, where only a few annotated samples of novel classes are needed with abundant samples of base classes. To this end, we propose Prototypical VoteNet to recognize and localize novel instances, which incorporates two new modules: Prototypical Vote Module (PVM) and Prototypical Head Module (PHM). Specifically, as the 3D basic geometric structures can be shared among categories, PVM is designed to leverage class-agnostic geometric prototypes, which are learned from base classes, to refine local features of novel categories. Then PHM is proposed to utilize class prototypes to enhance the global feature of each object, facilitating subsequent object localization and classification, which is trained by the episodic training strategy. To evaluate the model in this new setting, we contribute two new benchmark datasets, FS-ScanNet and FS-SUNRGBD. We conduct extensive experiments to demonstrate the effectiveness of Prototypical VoteNet, and our proposed method shows significant and consistent improvements compared to baselines on two benchmark datasets.

In this paper, we propose a geometric neural network with edge-aware refinement (GeoNet++) to jointly predict both depth and surface normal maps from a single image. Building on top of two-stream CNNs, GeoNet++ captures the geometric relationships between depth and surface normals with the proposed depth-to-normal and normal-to-depth modules. In particular, the "depth-to-normal" module exploits the least square solution of estimating surface normals from depth to improve their quality, while the "normal-to-depth" module refines the depth map based on the constraints on surface normals through kernel regression. Boundary information is exploited via an edge-aware refinement module. GeoNet++ effectively predicts depth and surface normals with high 3D consistency and sharp boundaries resulting in better reconstructed 3D scenes. Note that GeoNet++ is generic and can be used in other depth/normal prediction frameworks to improve 3D reconstruction quality and pixel-wise accuracy of depth and surface normals. Furthermore, we propose a new 3D geometric metric (3DGM) for evaluating depth prediction in 3D. In contrast to current metrics that focus on evaluating pixel-wise error/accuracy, 3DGM measures whether the predicted depth can reconstruct high quality 3D surface normals. This is a more natural metric for many 3D application domains. Our experiments on NYUD-V2 [1] and KITTI [2] datasets verify that GeoNet++ produces fine boundary details and the predicted depth can be used to reconstruct high quality 3D surfaces.

In 2D image processing, some attempts decompose images into high and low frequency components for describing edge and smooth parts respectively. Similarly, the contour and flat area of 3D objects, such as the boundary and seat area of a chair, describe different but also complementary geometries. However, such investigation is lost in previous deep networks that understand point clouds by directly treating all points or local patches equally. To solve this problem, we propose Geometry-Disentangled Attention Network (GDANet). GDANet introduces Geometry-Disentangle Module to dynamically disentangle point clouds into the contour and flat part of 3D objects, respectively denoted by sharp and gentle variation components. Then GDANet exploits Sharp-Gentle Complementary Attention Module that regards the features from sharp and gentle variation components as two holistic representations, and pays different attentions to them while fusing them respectively with original point cloud features. In this way, our method captures and refines the holistic and complementary 3D geometric semantics from two distinct disentangled components to supplement the local information. Extensive experiments on 3D object classification and segmentation benchmarks demonstrate that GDANet achieves the state-of-the-arts with fewer parameters.

In this paper, we present a conceptually simple, strong, and efficient framework for panoptic segmentation, called Panoptic FCN. Our approach aims to represent and predict foreground things and background stuff in a unified fully convolutional pipeline. In particular, Panoptic FCN encodes each object instance or stuff category into a specific kernel weight with the proposed kernel generator and produces the prediction by convolving the high-resolution feature directly. With this approach, instance-aware and semantically consistent prosperties for things and stuff can be respectively satisfied in a simple generate-kernel-then-segment workflow. Without extra boxes for localization or instance separation, the proposed approach outperforms previous box-based and -free models with high efficiency on COCO, Cityscapes, and Mapillary Vistas datasets with single scale input. Our code is made publicly available at https://github.com/Jia-Research-Lab/PanopticFCN.1

Point cloud semantic segmentation often requires large-scale annotated training data, but clearly, point-wise labels are too tedious to prepare. While some recent methods propose to train a 3D network with small percentages of point labels, we take the approach to an extreme and propose "One Thing One Click," meaning that the annotator only needs to label one point per object. To leverage these extremely sparse labels in network training, we design a novel self-training approach, in which we iteratively conduct the training and label propagation, facilitated by a graph propagation module. Also, we adopt a relation network to generate the per-category prototype and explicitly model the similarity among graph nodes to generate pseudo labels to guide the iterative training. Experimental results on both ScanNet-v2 and S3DIS show that our self-training approach, with extremely-sparse annotations, outperforms all existing weakly supervised methods for 3D semantic segmentation by a large margin, and our results are also comparable to those of the fully supervised counterparts.

Indoor scene semantic parsing from RGB images is very challenging due to occlusions, object distortion, and view-point variations. Going beyond prior works that leverage geometry information, typically paired depth maps, we present a new approach, a 3D-to-2D distillation framework, that enables us to leverage 3D features extracted from large-scale 3D data repositories (e.g., ScanNet-v2) to enhance 2D features extracted from RGB images. Our work has three novel contributions. First, we distill 3D knowledge from a pretrained 3D network to supervise a 2D network to learn simulated 3D features from 2D features during the training, so the 2D network can infer without requiring 3D data. Second, we design a two-stage dimension normalization scheme to calibrate the 2D and 3D features for better integration. Third, we design a semantic-aware adversarial training model to extend our framework for training with un-paired 3D data. Extensive experiments on various datasets, ScanNet-V2, S3DIS, and NYU-v2, demonstrate the superiority of our approach. Also, experimental results show that our 3D-to-2D distillation improves the model generalization.

We introduce Position Adaptive Convolution (PAConv), a generic convolution operation for 3D point cloud processing. The key of PAConv is to construct the convolution kernel by dynamically assembling basic weight matrices stored in Weight Bank, where the coefficients of these weight matrices are self-adaptively learned from point positions through ScoreNet. In this way, the kernel is built in a data-driven manner, endowing PAConv with more flexibility than 2D convolutions to better handle the irregular and unordered point cloud data. Besides, the complexity of the learning process is reduced by combining weight matrices instead of brutally predicting kernels from point positions.Furthermore, different from the existing point convolution operators whose network architectures are often heavily engineered, we integrate our PAConv into classical MLP-based point cloud pipelines without changing network configurations. Even built on simple networks, our method still approaches or even surpasses the state-of-the-art models, and significantly improves baseline performance on both classification and segmentation tasks, yet with decent efficiency. Thorough ablation studies and visualizations are provided to understand PAConv. Code is released on https://github.com/CVMI-Lab/PAConv.

We present a new domain adaptive self-training pipeline, named ST3D, for unsupervised domain adaptation on 3D object detection from point clouds. First, we pre-train the 3D detector on the source domain with our proposed random object scaling strategy for mitigating the negative effects of source domain bias. Then, the detector is iteratively improved on the target domain by alternatively conducting two steps, which are the pseudo label updating with the developed quality-aware triplet memory bank and the model training with curriculum data augmentation. These specific designs for 3D object detection enable the detector to be trained with consistent and high-quality pseudo labels and to avoid overfitting to the large number of easy examples in pseudo labeled data. Our ST3D achieves state-of-the-art performance on all evaluated datasets and even surpasses fully supervised results on KITTI 3D object detection benchmark. Code will be available at https://github.com/CVMI-Lab/ST3D.

While self-training has advanced semi-supervised semantic segmentation, it severely suffers from the long-tailed class distribution on real-world semantic segmentation datasets that make the pseudo-labeled data bias toward majority classes. In this paper, we present a simple and yet effective Distribution Alignment and Random Sampling (DARS) method to produce unbiased pseudo labels that match the true class distribution estimated from the labeled data. Besides, we also contribute a progressive data augmentation and labeling strategy to facilitate model training with pseudo-labeled data. Experiments on both Cityscapes and PASCAL VOC 2012 datasets demonstrate the effectiveness of our approach. Albeit simple, our method performs favorably in comparison with stateof-the-art approaches. Code will be available at https://github.com/CVMI-Lab/DARS.

Aggregating features from different depths of a network is widely adopted to improve the network capability. Lots of modern architectures are equipped with skip connections, which actually makes the feature aggregation happen in all these networks. Since different features tell different semantic meanings, there are inconsistencies and incompatibilities to be solved. However, existing works naively blend deep features via element-wise summation or concatenation with a convolution behind. Better feature aggregation method beyond summation or concatenation is rarely explored. In this paper, given two layers of features to be aggregated together, we first detect and identify where and what needs to be updated in one layer, then replace the feature at the identified location with the information of the other layer This process, which we call DEtect-rePLAce (DEPLA), enables us to avoid inconsistent patterns while keeping useful information in the merged outputs. Experimental results demonstrate our method largely boosts multiple baselines e.g. ResNet, FishNet and FPN on three major vision tasks including ImageNet classification, MS COCO object detection and instance segmentation.

Semantic segmentation is a challenging task that needs to handle large scale variations, deformations, and different viewpoints. In this paper, we develop a novel network named Gated Path Selection Network (GPSNet), which aims to adaptively select receptive fields while maintaining the dense sampling capability. In GPSNet, we first design a two-dimensional SuperNet, which densely incorporates features from growing receptive fields. And then, a Comparative Feature Aggregation (CFA) module is introduced to dynamically aggregate discriminative semantic context. In contrast to previous works that focus on optimizing sparse sampling locations on regular grids, GPSNet can adaptively harvest free form dense semantic context information. The derived adaptive receptive fields and dense sampling locations are data-dependent and flexible which can model various contexts of objects. On two representative semantic segmentation datasets, i.e., Cityscapes and ADE20K, we show that the proposed approach consistently outperforms previous methods without bells and whistles.

The automatic diagnosis of various conventional ophthalmic diseases from fundus images is important in clinical practice. However, developing such automatic solutions is challenging due to the requirement of a large amount of training data and the expensive annotations for medical images. This paper presents a novel self-supervised learning framework for retinal disease diagnosis to reduce the annotation efforts by learning the visual features from the unlabeled images. To achieve this, we present a rotation-oriented collaborative method that explores rotation-related and rotation-invariant features, which capture discriminative structures from fundus images and also explore the invariant property used for retinal disease classification. We evaluate the proposed method on two public benchmark datasets for retinal disease classification. The experimental results demonstrate that our method outperforms other self-supervised feature learning methods (around 4.2% area under the curve (AUC)). With a large amount of unlabeled data available, our method can surpass the supervised baseline for pathologic myopia (PM) and is very close to the supervised baseline for age-related macular degeneration (AMD), showing the potential benefit of our method in clinical practice.

We focus on Unsupervised Domain Adaptation (UDA) for the task of semantic segmentation. Recently, adversarial alignment has been widely adopted to match the marginal distribution of feature representations across two domains globally. However, this strategy fails in adapting the representations of the tail classes or small objects for semantic segmentation since the alignment objective is dominated by head categories or large objects. In contrast to adversarial alignment, we propose to explicitly train a domain-invariant classifier by generating and defensing against pointwise feature space adversarial perturbations. Specifically, we firstly perturb the intermediate feature maps with several attack objectives (i.e., discriminator and classifier) on each individual position for both domains, and then the classifier is trained to be invariant to the perturbations. By perturbing each position individually, our model treats each location evenly regardless of the category or object size and thus circumvents the aforementioned issue. Moreover, the domain gap in feature space is reduced by extrapolating source and target perturbed features towards each other with attack on the domain discriminator. Our approach achieves the state-of-the-art performance on two challenging domain adaptation tasks for semantic segmentation: GTA5 → Cityscapes and SYNTHIA → Cityscapes.

The goal of our paper is to semantically edit parts of an image matching a given text that describes desired attributes (e.g., texture, colour, and background), while preserving other contents that are irrelevant to the text. To achieve this, we propose a novel generative adversarial network (ManiGAN), which contains two key components: text-image affine combination module (ACM) and detail correction module (DCM). The ACM selects image regions relevant to the given text and then correlates the regions with corresponding semantic words for effective manipulation. Meanwhile, it encodes original image features to help reconstruct text-irrelevant contents. The DCM rectifies mismatched attributes and completes missing contents of the synthetic image. Finally, we suggest a new metric for evaluating image manipulation results, in terms of both the generation of new attributes and the reconstruction of text-irrelevant contents. Extensive experiments on the CUB and COCO datasets demonstrate the superior performance of the proposed method.

We present an end-to-end network to bridge the gap between training and inference pipeline for panoptic segmentation, a task that seeks to partition an image into semantic regions for "stuff" and object instances for "things". In contrast to recent works, our network exploits a parametrised, yet lightweight panoptic segmentation submodule, powered by an end-to-end learnt dense instance affinity, to capture the probability that any pair of pixels belong to the same instance. This panoptic submodule gives rise to a novel propagation mechanism for panoptic logits and enables the network to output a coherent panoptic segmentation map for both "stuff" and "thing" classes, without any post-processing. Reaping the benefits of end-to-end training, our full system sets new records on the popular street scene dataset, Cityscapes, achieving 61.4 PQ with a ResNet-50 backbone using only the fine annotations. On the challenging COCO dataset, our ResNet-50-based network also delivers state-of-the-art accuracy of 43.4 PQ. Moreover, our network flexibly works with and without object mask cues, performing competitively under both settings, which is of interest for applications with computation budgets.

Generative Adversarial Networks (GANs) can generate realistic fake face images that can easily fool human beings. On the contrary, a common Convolutional Neural Network(CNN) discriminator can achieve more than99.9%accuracyin discerning fake/real images. In this paper, we conduct an empirical study on fake/real faces, and have two important observations: firstly, the texture of fake faces is substantially different from real ones; secondly, global texture statistics are more robust to image editing and transferable to fake faces from different GANs and datasets. Motivated by the above observations, we propose a new architecture coined as Gram-Net, which leverages global image texture representations for robust fake image detection. Experimental results on several datasets demonstrate that our Gram-Netoutperforms existing approaches. Especially, our Gram-Netis more robust to image editings, e.g. down-sampling, JPEGcompression, blur, and noise. More importantly, our Gram-Net generalizes significantly better in detecting fake faces from GAN models not seen in the training phase and can perform decently in detecting fake natural images.

We propose a novel lightweight generative adversarial network for efficient image manipulation using natural language descriptions. To achieve this, a new word-level discriminator is proposed, which provides the generator with fine-grained training feedback at word-level, to facilitate training a lightweight generator that has a small number of parameters, but can still correctly focus on specific visual attributes of an image, and then edit them without affecting other contents that are not described in the text. Furthermore, thanks to the explicit training signal related to each word, the discriminator can also be simplified to have a lightweight structure. Compared with the state of the art, our method has a much smaller number of parameters, but still achieves a competitive manipulation performance. Extensive experimental results demonstrate that our method can better disentangle different visual attributes, then correctly map them to corresponding semantic words, and thus achieve a more accurate image modification using natural language descriptions.

A future video is the 2D projection of a 3D scene with predicted camera and object motion. Accurate future video prediction inherently requires understanding of 3D motion and geometry of a scene. In this paper, we propose a RGBD scene forecasting model with 3D motion decomposition. We predict ego-motion and foreground motion that are combined to generate a future 3D dynamic scene, which is then projected into a 2D image plane to synthesize future motion, RGB images and depth maps. Optional semantic maps can be integrated. Experimental results on KITTI and Driving datasets show that our model outperforms other state-of-the- arts in forecasting future RGBD dynamic scenes.

Adaptive inference is a promising technique to improve the computational efficiency of deep models at test time. In contrast to static models which use the same computation graph for all instances, adaptive networks can dynamically adjust their structure conditioned on each input. While existing research on adaptive inference mainly focuses on designing more advanced architectures, this paper investigates how to train such networks more effectively. Specifically, we consider a typical adaptive deep network with multiple intermediate classifiers. We present three techniques to improve its training efficacy from two aspects: 1) a Gradient Equilibrium algorithm to resolve the conflict of learning of different classifiers; 2) an Inline Subnetwork Collaboration approach and a One-for-all Knowledge Distillation algorithm to enhance the collaboration among classifiers. On multiple datasets (CIFAR-10, CIFAR-100 and ImageNet), we show that the proposed approach consistently leads to further improved efficiency on top of state-of-the-art adaptive deep networks.

Most video object segmentation approaches process objects separately. This incurs high computational cost when multiple objects exist. In this paper, we propose AGSS-VOS to segment multiple objects in one feed-forward path via instance-agnostic and instance-specific modules. Information from the two modules is fused via an attention-guided decoder to simultaneously segment all object instances in one path. The whole framework is end-to-end trainable with instance IoU loss. Experimental results on Youtube- VOS and DAVIS-2017 dataset demonstrate that AGSS-VOS achieves competitive results in terms of both accuracy and efficiency.

In this paper, we propose a novel controllable text-to-image generative adversarial network (ControlGAN), which can effectively synthesise high-quality images and also control parts of the image generation according to natural language descriptions. To achieve this, we introduce a word-level spatial and channel-wise attention-driven generator that can disentangle different visual attributes, and allow the model to focus on generating and manipulating subregions corresponding to the most relevant words. Also, a word-level discriminator is proposed to provide fine-grained supervisory feedback by correlating words with image regions, facilitating training an effective generator which is able to manipulate specific visual attributes without affecting the generation of other content. Furthermore, perceptual loss is adopted to reduce the randomness involved in the image generation, and to encourage the generator to manipulate specific attributes required in the modified text. Extensive experiments on benchmark datasets demonstrate that our method outperforms existing state of the art, and is able to effectively manipulate synthetic images using natural language descriptions.

We address the problem of image segmentation from natural language descriptions. Existing deep learning-based methods encode image representations based on the output of the last convolutional layer. One general issue is that the resulting image representation lacks multi-scale semantics, which are key components in advanced segmentation systems. In this paper, we utilize the feature pyramids inherently existing in convolutional neural networks to capture the semantics at different scales. To produce suitable information flow through the path of feature hierarchy, we propose Recurrent Refinement Network (RRN) that takes pyramidal features as input to refine the segmentation mask progressively. Experimental results on four available datasets show that our approach outperforms multiple baselines and state-of-the-art1.

We present a semi-parametric approach to photographic image synthesis from semantic layouts. The approach combines the complementary strengths of parametric and nonparametric techniques. The nonparametric component is a memory bank of image segments constructed from a training set of images. Given a novel semantic layout at test time, the memory bank is used to retrieve photographic references that are provided as source material to a deep network. The synthesis is performed by a deep network that draws on the provided photographic material. Experiments on multiple semantic segmentation datasets show that the presented approach yields considerably more realistic images than recent purely parametric techniques.

In this paper, we propose Geometric Neural Network (GeoNet) to jointly predict depth and surface normal maps from a single image. Building on top of two-stream CNNs, our GeoNet incorporates geometric relation between depth and surface normal via the new depth-to-normal and normal-to-depth networks. Depth-to-normal network exploits the least square solution of surface normal from depth and improves its quality with a residual module. Normal-to-depth network, contrarily, refines the depth map based on the constraints from the surface normal through a kernel regression module, which has no parameter to learn. These two networks enforce the underlying model to efficiently predict depth and surface normal for high consistency and corresponding accuracy. Our experiments on NYU v2 dataset verify that our GeoNet is able to predict geometrically consistent depth and normal maps. It achieves top performance on surface normal estimation and is on par with state-of-the-art depth estimation methods.

In this paper, we propose a self-boosted intelligent system for joint sign language recognition and automatic education. A novel Spatial-Temporal Net (ST-Net) is designed to exploit the temporal dynamics of localized hands for sign language recognition. Features from ST-Net can be deployed by our education system to detect failure modes of the learners. Moreover, the education system can help collect a vast amount of data for training ST-Net. Our sign language recognition and education system help improve each other step-by-step.On the one hand, benefited from accurate recognition system, the education system can detect the failure parts of the learner more precisely. On the other hand, with more training data gathered from the education system, the recognition system becomes more robust and accurate. Experiments on Hong Kong sign language dataset containing 227 commonly used words validate the effectiveness of our joint recognition and education system.

In this paper, we propose a generative multi-column network for image inpainting. This network synthesizes different image components in a parallel manner within one stage. To better characterize global structures, we design a confidence-driven reconstruction loss while an implicit diversified MRF regularization is adopted to enhance local details. The multi-column network combined with the reconstruction and MRF loss propagates local and global information derived from context to the target inpainting regions. Extensive experiments on challenging street view, face, natural objects and scenes manifest that our method produces visual compelling results even without previously common post-processing.

Liver cancer is one of the leading causes of cancer death. To assist doctors in hepatocellular carcinoma diagnosis and treatment planning, an accurate and automatic liver and tumor segmentation method is highly demanded in the clinical practice. Recently, fully convolutional neural networks (FCNs), including 2-D and 3-D FCNs, serve as the backbone in many volumetric image segmentation. However, 2-D convolutions cannot fully leverage the spatial information along the third dimension while 3-D convolutions suffer from high computational cost and GPU memory consumption. To address these issues, we propose a novel hybrid densely connected UNet (H-DenseUNet), which consists of a 2-D DenseUNet for efficiently extracting intra-slice features and a 3-D counterpart for hierarchically aggregating volumetric contexts under the spirit of the auto-context algorithm for liver and tumor segmentation. We formulate the learning process of the H-DenseUNet in an end-to-end manner, where the intra-slice representations and inter-slice features can be jointly optimized through a hybrid feature fusion layer. We extensively evaluated our method on the data set of the MICCAI 2017 Liver Tumor Segmentation Challenge and 3DIRCADb data set. Our method outperformed other state-of-the-arts on the segmentation results of tumors and achieved very competitive performance for liver segmentation even with a single model.

Scene parsing is challenging for unrestricted open vocabulary and diverse scenes. In this paper, we exploit the capability of global context information by different-region-based context aggregation through our pyramid pooling module together with the proposed pyramid scene parsing network (PSPNet). Our global prior representation is effective to produce good quality results on the scene parsing task, while PSPNet provides a superior framework for pixel-level prediction. The proposed approach achieves state-of-the-art performance on various datasets. It came first in ImageNet scene parsing challenge 2016, PASCAL VOC 2012 benchmark and Cityscapes benchmark. A single PSPNet yields the new record of mIoU accuracy 85.4% on PASCAL VOC 2012 and accuracy 80.2% on Cityscapes.

RGBD semantic segmentation requires joint reasoning about 2D appearance and 3D geometric information. In this paper we propose a 3D graph neural network (3DGNN) that builds a k-nearest neighbor graph on top of 3D point cloud. Each node in the graph corresponds to a set of points and is associated with a hidden representation vector initialized with an appearance feature extracted by a unary CNN from 2D images. Relying on recurrent functions, every node dynamically updates its hidden representation based on the current status and incoming messages from its neighbors. This propagation model is unrolled for a certain number of time steps and the final per-node representation is used for predicting the semantic class of each pixel. We use back-propagation through time to train the model. Extensive experiments on NYUD2 and SUN-RGBD datasets demonstrate the effectiveness of our approach.

The goal of connectomics is to manifest the interconnections of neural system with the Electron Microscopy (EM) images. However, the formidable size of EM image data renders human annotation impractical, as it may take decades to fulfill the whole job. An alternative way to reconstruct the connectome can be attained with the computerized scheme that can automatically segment the neuronal structures. The segmentation of EM images is very challenging as the depicted structures can be very diverse.To address this difficult problem, a deep contextual network is proposed here by leveraging multi-level contextual information from the deep hierarchical structure to achieve better segmentation performance.To further improve the robustness against the vanishing gradients and strengthen the capability of the back-propagation of gradient flow, auxiliary classifiers are incorporated in the architecture of our deep neural network. It will be shown that our method can effectively parse the semantic meaning from the images with the underlying neural network and accurately delineate the structural boundaries with the reference of low-level contextual cues. Experimental results on the benchmark dataset of 2012 ISBI segmentation challenge of neuronal structures suggest that the proposed method can outperform the state-of-the-art methods by a large margin with respect to different evaluation measurements. Our method can potentially facilitate the automatic connectome analysis from EM images with less human intervention effort.

The morphology of glands has been used routinely by pathologists to assess the malignancy degree of adenocarcinomas. Accurate segmentation of glands from histology images is a crucial step to obtain reliable morphological statistics for quantitative diagnosis. In this paper, we proposed an efficient deep contour-aware network (DCAN) to solve this challenging problem under a unified multi-task learning framework. In the proposed network, multi-level contextual features from the hierarchical architecture are explored with auxiliary supervision for accurate gland segmentation. When incorporated with multi-task regularization during the training, the discriminative capability of intermediate features can be further improved. Moreover, our network can not only output accurate probability maps of glands, but also depict clear contours simultaneously for separating clustered objects, which further boosts the gland segmentation performance. This unified framework can be efficient when applied to large-scale histopathological data without resorting to additional steps to generate contours based on low-level cues for post-separating. Our method won the 2015 MICCAI Gland Segmentation Challenge out of 13 competitive teams, surpassing all the other methods by a significant margin.

Aiming at simultaneous detection and segmentation (SD-S), we propose a proposal-free framework, which detect and segment object instances via mid-level patches. We design a unified trainable network on patches, which is followed by a fast and effective patch aggregation algorithm to infer object instances. Our method benefits from end-to-end training. Without object proposal generation, computation time can also be reduced. In experiments, our method yields results 62.1% and 61.8% in terms of mAPr on VOC2012 segmentation val and VOC2012 SDS val, which are state-of-the-art at the time of submission. We also report results on Microsoft COCO test-std/test-dev dataset in this paper.

We propose an object clique potential for semantic segmentation. Our object clique potential addresses the misclassified object-part issues arising in solutions based on fully-convolutional networks. Our object clique set, compared to that yielded from segment-proposal-based approaches, is with a significantly smaller size, making our method consume notably less computation. Regarding system design and model formation, our object clique potential can be regarded as a functional complement to local-appearance-based CRF models and works in synergy with these effective approaches for further performance improvement. Extensive experiments verify our method.