This paper concentrates on Multi-modal Referring Video Segmentation task, where a well optimized model is able to recognize and segment the target objects referred by the given guidance signals, e.g., language description. Early approaches model this task as a sequence prediction problem. The lack of a global view of video content leads to difficulties in effectively utilizing inter-frame relationships. Some recent works propose to perform temporal modeling with vanilla attention mechanism. However, the condensed visual representation tends to be messy about target information due to occlusion or motion blur. Unlimited non-local operation would spread such noise to all the sequences and interfere with the extraction of global representations. To address the above issue, we present Semantic-assisted Object Cluster network (SOC) and the improved SOC++ in this paper. Our method unifies temporally selective interaction and cross-modal alignment to achieve video-level understanding. In SOC++, a proxy-assisted multi-modal fusion module is introduced to perform preliminary bidirectional activation. Then a semantic integration module with progressive frame-to-video structure facilitates joint space learning across modalities and time steps. Considering that potential noisy visual embeddings would impair the overall representation of target objects in unconstrained inter-frame interactions, we propose to perform tendentious video aggregation through emphasizing the indicative role of the informative frames with lower entropy in this part. A multi-modal query contrastive supervision is also utilized to help construct well-aligned joint space at the video level. Moreover, to integrate the advantage of high-level video information and the low-level details of each frame, we introduce a dynamic query fusion module that performs joint updating of these embeddings. We conduct extensive experiments on popular referring video segmentation benchmarks, and our method outperforms state-of-the-art competitors on all benchmarks by a remarkable margin. Besides, the emphasis on temporal coherence enhances the segmentation stability and adaptability of our method in processing text expressions with temporal variations..

Arbitrary-scale super-resolution is an essential image upsampling task, typically tackled with implicit neural representations. Unlike INR-based methods that rely on slow per-pixel decoding, Gaussian Splatting is promising for arbitrary-scale super-resolution when the explicit region-based nature of GS allows for highly efficient rendering via a lightweight decoder. Recently, Gaussian splatting has outperformed implicit neural methods in 3D scenes. However, existing attempts to address the super-resolution problem with Gaussian splatting face efficiency and accuracy challenges, such as redundant Gaussian primitives and discrete pixel sampling. Efficient Gaussian applications may result in continuous texture constraints due to limited feature richness in explicit fields, particularly the mismatch between the learned Gaussian fields and out-of-distribution sampling rates. Moreover, insufficient discrete sampling based on the given upscale factor may fail to accurately represent the splatted Gaussian field in screen space, causing high-frequency signal redundancy and aliasing. To address these challenges, we propose a compact Gaussian splatting method for efficient arbitrary-scale super-resolution, CGSSR. It constructs an efficient Gaussian space by distilling the image into a reduced number of primitives, each represented by a compact, low-dimensional feature embedding. To balance detail preservation and anti-aliasing, we introduce a scale-aware smoothing filter to regulate splatting frequency. Extensive experiments show that CGSSR achieves superior performance over existing Gaussian-based methods, especially at large scales, with higher efficiency.

Pose-guided human image generation aims to render a source image in a specific pose. Current methods predominantly employ 2D-based signals, which exhibit inherent information deficits, as pose conditions. This leads to difficulty in establishing precise source-target appearance-pose correspondence and further causing uncertainty in predicting self-occluded regions’ appearance. To address these issues, we propose a 3D Pose Conditional Diffusion model (3DPCD) that leverages a human parametric model to integrate comprehensive and adjustable 3D control into forward–backward diffusion steps. Specifically, we employ Fourier-transformed SMPL-X as the 3D pose representation to facilitate precise source-target correspondence by understanding the complete pose information. Building on this, we further propose a hierarchical appearance-pose alignment method, which aligns appearance with the complete pose information at both global and local levels. Moreover, motivated by the fact that human pose transformation is a progressive process in 3D space and our 3D pose representation is adjustable, we integrate progressively interpolated 3D control into a series of sampling steps. This effectively mitigates uncertainties in pixel transfer between poses. It should be noted that the proposed explicit pose-guided strategy also supports flexible adjustment of pose, shape, and viewpoint. Both quantitative and qualitative evaluations demonstrate that our 3DPCD outperforms state-of-the-art methods on the widely used DeepFashion InShop benchmark and our newly constructed PoseWeb-33 dataset, which features richer appearance variations and more diverse conditional poses.

Zero-shot Referring Image Segmentation (RIS) identifies the instance mask that best aligns with a specified referring expression without training and fine-tuning, significantly reducing the labor-intensive annotation process. Despite achieving commendable results, previous CLIP-based models have a critical drawback: the models exhibit a notable reduction in their capacity to discern relative spatial relationships of objects. This is because they generate all possible masks on an image and evaluate each masked region for similarity to the given expression, often resulting in decreased sensitivity to direct positional clues in text inputs. Moreover, most methods have weak abilities to manage relationships between primary words and their contexts, causing confusion and reduced accuracy in identifying the correct target region. To address these challenges, we propose IteRPrimE (Iterative Grad-CAM Refinement and Primary word Emphasis), which leverages a saliency heatmap through Grad-CAM from a Vision-Language Pre-trained (VLP) model for image-text matching. An iterative Grad-CAM refinement strategy is introduced to progressively enhance the model's focus on the target region and overcome positional insensitivity, creating a self-correcting effect. Additionally, we design the Primary Word Emphasis module to help the model handle complex semantic relations, enhancing its ability to attend to the intended object. Extensive experiments conducted on the RefCOCO/+/g, and PhraseCut benchmarks demonstrate that IteRPrimE outperforms previous SOTA zero-shot methods, particularly excelling in out-of-domain scenarios.

Multimodal language models (MLLMs) are increasingly being applied in real-world environments, necessitating their ability to interpret 3D spaces and comprehend temporal dynamics. Current methods often rely on specialized architectural designs or task-specific fine-tuning to achieve this. We introduce Coarse Correspondences, a simple lightweight method that enhances MLLMs’ spatial-temporal reasoning with 2D images as input, without modifying the architecture or requiring task-specific fine-tuning. Our method uses a lightweight tracking model to identify primary object correspondences between frames in a video or across different image viewpoints, and then conveys this information to MLLMs through visual prompting. We demonstrate that this simple training-free approach brings substantial gains to GPT4-V/O consistently on four benchmarks that require spatial-temporal reasoning, including +20.5% improvement on ScanQA, +9.7% on OpenEQA’s episodic memory subset, +6.0% on the long-form video benchmark EgoSchema, and +11% on the R2R navigation benchmark. Additionally, we show that Coarse Correspondences can also enhance open-source MLLMs’ spatial reasoning (by +6.9% on ScanQA) when applied in both training and inference and that the improvement can generalize to unseen datasets such as SQA3D (+3.1%). Taken together, we show that Coarse Correspondences effectively and efficiently boosts models’ performance on downstream tasks requiring spatial-temporal reasoning.

Reference Audio-Visual Segmentation (Ref-AVS) aims to provide a pixel-wise scene understanding in Language-aided Audio-Visual Scenes (LAVS). This task requires the model to continuously segment objects referred to by text and audio from a video. Previous dual-modality methods always fail due to the lack of a third modality and the existing triple-modality method struggles with spatio-temporal consistency, leading to the target shift of different frames. In this work, we introduce a novel framework, termed SAM2-LOVE, which integrates textual, audio, and visual representations into a learnable token to prompt and align SAM2 for achieving Ref-AVS in the LAVS. Technically, our approach includes a multimodal fusion module aimed at improving multimodal understanding of SAM2, as well as token propagation and accumulation strategies designed to enhance spatio-temporal consistency without forgetting historical information. We conducted extensive experiments to demonstrate that SAM2-LOVE outperforms the SOTA by 8.5% in $\mathcal{J}\& \mathcal{F}$ on the Ref-AVS benchmark and showcase the simplicity and effectiveness of the components. Our code will be available here.

Large Vision Language Models (LVLMs) have achieved significant success across multi-modal tasks. However, the computational cost of processing long visual tokens can be prohibitively expensive on resource-limited devices. Previous methods have identified redundancy in visual tokens within the Large Language Model (LLM) decoder layers and have mitigated this by pruning tokens using a predefined or fixed ratio, thereby reducing computational over-head. Nonetheless, we observe that the impact of pruning ratio varies across different LLM layers and instances (image-prompt pairs). Therefore, it is essential to develop a layer-wise and instance-wise vision token pruning strategy to balance computational cost and model performance effectively. We propose ATP-LLaVA, a novel approach that adaptively determines instance-specific token pruning ratios for each LLM layer. Specifically, we introduce an Adaptive Token Pruning (ATP) module, which computes the importance score and pruning threshold based on input instance adaptively. The ATP module can be seamlessly integrated between any two LLM layers with negligible computational overhead. Additionally, we develop a Spatial Augmented Pruning (SAP) strategy that prunes visual tokens with both token redundancy and spatial modeling perspectives. Our approach reduces the average token count by 75% while maintaining performance, with only a minimal 1.9% degradation across seven widely used benchmarks.

Vision-Language Models (VLMs) have achieved remarkable success in various multi-modal tasks, but they are often bottlenecked by the limited context window and high computational cost of processing high-resolution image inputs and videos. Vision compression can alleviate this problem by reducing the vision token count. Previous approaches compress vision tokens with external modules and force LLMs to understand the compressed ones, leading to visual information loss. However, the LLMs’ understanding paradigm of vision tokens is not fully utilised in the compression learning process. We propose VoCo-LLaMA, the first approach to compress vision tokens using LLMs. By introducing Vision Compression tokens during the vision instruction tuning phase and leveraging attention distillation, our method distill how LLMs comprehend vision tokens into their processing of VoCo tokens. VoCo-LLaMA facilitates effective vision compression and improves the computational efficiency during the inference stage. Specifically, our method can achieve a 576× compression rate while maintaining 83.7% performance. Furthermore, through continuous training using time-series compressed token sequences of video frames, VoCo-LLaMA demonstrates the ability to understand temporal correlations, outperforming previous methods on popular video question-answering benchmarks. Our approach presents a promising way to unlock the full potential of VLMs’ contextual window, enabling more scalable multi-modal applications. The project page can be accessed via https://yxxxb.github.io/VoCo-LLaMA-page.

Recent advancements in diffusion frameworks have significantly enhanced video editing, achieving high fidelity and strong alignment with textual prompts. However, conventional approaches using image diffusion models fall short in handling video dynamics, particularly for challenging temporal edits like motion adjustments. While current video diffusion models produce high-quality results, adapting them for efficient editing remains difficult due to the heavy computational demands that prevent the direct application of previous image editing techniques. To overcome these limitations, we introduce FADE—a training-free yet highly effective video editing approach that fully leverages the inherent priors from pre-trained video diffusion models via frequency-aware factorization. Rather than simply using these models, we first analyze the attention patterns within the video model to reveal how video priors are distributed across different components. Building on these insights, we propose a factorization strategy to optimize each component’s specialized role. Furthermore, we devise spectrum-guided modulation to refine the sampling trajectory with frequency domain cues, preventing information leakage and supporting efficient, versatile edits while preserving the basic spatial and temporal structure. Extensive experiments on real-world videos demonstrate that our method consistently delivers high-quality, realistic and temporally coherent editing results both qualitatively and quantitatively. Code is available at https://github.com/EternalEvan/FADE.

3D Gaussian Splatting has demonstrated notable success in large-scale scene reconstruction, but challenges persist due to high training memory consumption and storage overhead. Hybrid representations that integrate implicit and explicit features offer a way to mitigate these limitations. However, when applied in parallelized block-wise training, two critical issues arise since reconstruction accuracy deteriorates due to reduced data diversity when training each block independently, and parallel training restricts the number of divided blocks to the available number of GPUs. To address these issues, we propose Momentum-GS, a novel approach that leverages momentum-based self-distillation to promote consistency and accuracy across the blocks while decoupling the number of blocks from the physical GPU count. Our method maintains a teacher Gaussian decoder updated with momentum, ensuring a stable reference during training. This teacher provides each block with global guidance in a self-distillation manner, promoting spatial consistency in reconstruction. To further ensure consistency across the blocks, we incorporate block weighting, dynamically adjusting each block's weight according to its reconstruction accuracy. Extensive experiments on large-scale scenes show that our method consistently outperforms existing techniques, achieving a 12.8% improvement in LPIPS over CityGaussian with much fewer divided blocks and establishing a new state of the art. Project page: https://jixuan-fan.github.io/Momentum-GS_Page/

Joint reconstruction of human-object interaction marks a significant milestone in comprehending the intricate interrelations between humans and their surrounding environment. Nevertheless, previous optimization methods often struggle to achieve physically plausible reconstruction results due to the lack of prior knowledge about human-object interactions. In this paper, we introduce ScoreHOI, an effective diffusion-based optimizer that introduces diffusion priors for the precise recovery of human-object interactions. By harnessing the controllability within score-guided sampling, the diffusion model can reconstruct a conditional distribution of human and object pose given the image observation and object feature. During inference, the ScoreHOI effectively improves the reconstruction results by guiding the denoising process with specific physical constraints. Furthermore, we propose a contact-driven iterative refinement approach to enhance the contact plausibility and improve the reconstruction accuracy. Extensive evaluations on standard benchmarks demonstrate ScoreHOI's superior performance over state-of-the-art methods, highlighting its ability to achieve a precise and robust improvement in joint human-object interaction reconstruction.

Open-vocabulary segmentation aims to achieve segmentation of arbitrary categories given unlimited text inputs as guidance. To achieve this, recent works have focused on developing various technical routes to exploit the potential of large-scale pre-trained vision-language models and have made significant progress on existing benchmarks. However, we find that existing test sets are limited in measuring the models' comprehension of ``open-vocabulary" concepts, as their semantic space closely resembles the training space, even with many overlapping categories. To this end, we present a new benchmark named OpenBench that differs significantly from the training semantics. It is designed to better assess the model's ability to understand and segment a wide range of real-world concepts. When testing existing methods on OpenBench, we find that their performance diverges from the conclusions drawn on existing test sets. In addition, we propose a method named OVSNet to improve the segmentation performance for diverse and open scenarios. Through elaborate fusion of heterogeneous features and cost-free expansion of the training space, OVSNet achieves state-of-the-art results on both existing datasets and our proposed OpenBench. Corresponding analysis demonstrate the soundness and effectiveness of our proposed benchmark and method.

Performing general language-conditioned bimanual manipulation tasks is of great importance for many applications ranging from household service to industrial assembly. However, collecting bimanual manipulation data is expensive due to the high-dimensional action space, which poses challenges for conventional methods to handle general bimanual manipulation tasks. In contrast, unimanual policy has recently demonstrated impressive generalizability across a wide range of tasks because of scaled model parameters and training data, which can provide sharable manipulation knowledge for bimanual systems. To this end, we propose a plug-and-play method named AnyBimanual, which transfers pre-trained unimanual policy to general bimanual manipulation policy with few bimanual demonstrations. Specifically, we first introduce a skill manager to dynamically schedule the skill representations discovered from pre-trained unimanual policy for bimanual manipulation tasks, which linearly combines skill primitives with task-oriented compensation to represent the bimanual manipulation instruction. To mitigate the observation discrepancy between unimanual and bimanual systems, we present a visual aligner to generate soft masks for visual embedding of the workspace, which aims to align visual input of unimanual policy model for each arm with those during pretraining stage. AnyBimanual shows superiority on 12 simulated tasks from RLBench2 with a sizable 12.67% improvement in success rate over previous methods. Experiments on 9 real-world tasks further verify its practicality with an average success rate of 84.62%.

Benefiting from the advances in large language models and cross-modal alignment, existing multimodal large language models have achieved prominent performance in image and short video understanding. However, the understanding of long videos is still challenging, as their long-context nature results in significant computational and memory overhead. Most existing work treats long videos in the same way as short videos, which is inefficient for real-world applications and hard to generalize to even longer videos. To address these issues, we propose Flash-VStream, an efficient video language model capable of processing extremely long videos and responding to user queries in real time. Particularly, we design a Flash Memory module, containing a low-capacity context memory to aggregate long-context temporal information and model the distribution of information density, and a high-capacity augmentation memory to retrieve detailed spatial information based on this distribution. Compared to existing models, Flash-VStream achieves significant reductions in inference latency. Extensive experiments on long video benchmarks and comprehensive video benchmarks, i.e., EgoSchema, MLVU, LVBench, MVBench and Video-MME, demonstrate the state-of-the-art performance and outstanding efficiency of our method. Code is available at https://github.com/IVGSZ/Flash-VStream.

Background consistency remains a significant challenge in image editing tasks. Despite extensive developments, existing works still face a trade-off between maintaining similarity to the original image and generating content that aligns with the target. Here, we propose KV-Edit, a training-free approach that uses KV cache in DiTs to maintain background consistency, where background tokens are preserved rather than regenerated, eliminating the need for complex mechanisms or expensive training, ultimately generating new content that seamlessly integrates with the background within user-provided regions. We further explore the memory consumption of the KV cache during editing and optimize the space complexity to $O(1)$ using an inversion-free method. Our approach is compatible with any DiT-based generative model without additional training. Experiments demonstrate that KV-Edit significantly outperforms existing approaches in terms of both background and image quality, even surpassing training-based methods. Project webpage is available at https://xilluill.github.io/projectpages/KV-Edit

Embodied Instruction Following (EIF) requires agents to complete human instruction by interacting objects in complicated surrounding environments. Conventional methods directly consider the sparse human instruction to generate action plans for agents, which usually fail to achieve human goals because of the instruction incoherence in action descriptions. On the contrary, we propose ThinkBot that reasons the thought chain in human instruction to recover the missing action descriptions, so that the agent can successfully complete human goals by following the coherent instruction. Specifically, we first design an instruction completer based on large language models to recover the missing actions with interacted objects between consecutive human instruction, where the perceived surrounding environments and the completed sub-goals are considered for instruction completion. Based on the partially observed scene semantic maps, we present an object localizer to infer the position of interacted objects and the related Bayesian uncertainty for close-loop planning. Extensive experiments in the simulated environment show that our ThinkBot outperforms the state-of-the-art EIF methods by a sizable margin in both success rate and execution efficiency. Project page: https://guanxinglu.github.io/thinkbot/.

Large language models (LLMs), such as GPT-4, have shown remarkable performance in natural language processing (NLP) tasks, including challenging mathematical reasoning. However, most existing open-source models are only pre-trained on large-scale internet data and without math-related optimization. In this paper, we present WizardMath, which enhances the mathematical reasoning abilities of LLMs, by applying our proposed Reinforcement Learning from Evol-Instruct Feedback (RLEIF) method to the domain of math. Through extensive experiments on two mathematical reasoning benchmarks, namely GSM8k and MATH, we reveal the extraordinary capabilities of our model. Remarkably, WizardMath-Mistral 7B surpasses all other open-source LLMs by a substantial margin. Furthermore, WizardMath 70B even outperforms ChatGPT-3.5, Claude Instant, Gemini Pro and Mistral Medium. Additionally, our preliminary exploration highlights the pivotal role of instruction evolution and process supervision in achieving exceptional math performance.

Image enhancement finds wide-ranging applications in real-world scenarios due to complex environments and the inherent limitations of imaging devices. Recent diffusion-based methods yield promising outcomes but necessitate prolonged and computationally intensive iterative sampling. In response, we propose InstaRevive, a straightforward yet powerful image enhancement framework that employs score-based diffusion distillation to harness potent generative capability and minimize the sampling steps. To fully exploit the potential of the pre-trained diffusion model, we devise a practical and effective diffusion distillation pipeline using dynamic noise control to address inaccuracies in updating direction during score matching. Our noise control strategy enables a dynamic diffusing scope, facilitating precise learning of denoising trajectories within the diffusion model and ensuring accurate distribution matching gradients during training. Additionally, to enrich guidance for the generative power, we incorporate textual prompts via image captioning as auxiliary conditions, fostering further exploration of the diffusion model. Extensive experiments substantiate the efficacy of our framework across a diverse array of challenging tasks and datasets, unveiling the compelling efficacy and efficiency of InstaRevive in delivering high-quality and visually appealing results.

Multi-task robotic bimanual manipulation is becoming increasingly popular as it enables sophisticated tasks that require diverse dual-arm collaboration patterns. Compared to unimanual manipulation, bimanual tasks pose challenges to understanding the multi-body spatiotemporal dynamics. An existing method ManiGaussian [30] pioneers encoding the spatiotemporal dynamics into the visual representation via Gaussian world model for single-arm settings, which ignores the interaction of multiple embodiments for dual-arm systems with significant performance drop. In this paper, we propose ManiGaussian++, an extension of ManiGaussian framework that improves multi-task bimanual manipulation by digesting multi-body scene dynamics through a hierarchical Gaussian world model. To be specific, we first generate task-oriented Gaussian Splatting from intermediate visual features, which aims to differentiate acting and stabilizing arms for multi-body spatiotemporal dynamics modeling. We then build a hierarchical Gaussian world model with the leader-follower architecture, where the multi-body spatiotemporal dynamics is mined for intermediate visual representation via future scene prediction. The leader predicts Gaussian Splatting deformation caused by motions of the stabilizing arm, through which the follower generates the physical consequences resulted from the movement of the acting arm. As a result, our method significantly outperforms the current state-of-the-art bimanual manipulation techniques by an improvement of 20.2% in 10 simulated tasks, and achieves 60% success rate on average in 9 challenging real-world tasks. Our code is available at https://github.com/April-Yz/ManiGaussian_Bimanual.

We introduce a model named DreamLight for universal image relighting in this work, which can seamlessly composite subjects into a new background while maintaining aesthetic uniformity in terms of lighting and color tone. The background can be specified by natural images (image-based relighting) or generated from unlimited text prompts (text-based relighting). Existing studies primarily focus on image-based relighting, while with scant exploration into text-based scenarios. Some works employ intricate disentanglement pipeline designs relying on environment maps to provide relevant information, which grapples with the expensive data cost required for intrinsic decomposition and light source. Other methods take this task as an image translation problem and perform pixel-level transformation with autoencoder architecture. While these methods have achieved decent harmonization effects, they struggle to generate realistic and natural light interaction effects between the foreground and background. To alleviate these challenges, we reorganize the input data into a unified format and leverage the semantic prior provided by the pretrained diffusion model to facilitate the generation of natural results. Moreover, we propose a Position-Guided Light Adapter (PGLA) that condenses light information from different directions in the background into designed light query embeddings, and modulates the foreground with direction-biased masked attention. In addition, we present a post-processing module named Spectral Foreground Fixer (SFF) to adaptively reorganize different frequency components of subject and relighted background, which helps enhance the consistency of foreground appearance. Extensive comparisons and user study demonstrate that our DreamLight achieves remarkable relighting performance.

3D Gaussian Splatting (3DGS) has emerged as a high-quality and efficient alternative to Neural Radiance Fields (NeRF), offering distinct advantages in scene representation and suitability for multi-view 3D object detection (MV-3DOD) tasks. However, conventional 3DGS-based detection methods typically employ a reconstruction-then-detection pipeline, which is time-consuming and unsuitable for real-time applications. This approach arises from 3DGS’s original design for reconstruction tasks, which lacks network-based training. In this paper, we propose GS-3Det, an online Gaussian detection framework for MV-3DOD, achieving real-time performance through a single forward pass for scene reconstruction and detection. Specifically, we introduce a detection-aware Gaussian grid that enables directly prediction of explicit 3D Gaussians from multi-view images, enabling efficient and robust 3D scene understanding. Additionally, we propose a Dual-Path Consistency module that leverages 3D constraints to improve the accuracy of Gaussian grid representation and detection. Experiments on the ScanNet V2 dataset demonstrate that GS-3Det surpasses state-of-the-art methods by 3.5% in [email protected] and 2.7% in [email protected], underscoring its generalization capability and real-time performance.

While significant progress has been achieved in single-person image generation, generating realistic images of multiple individuals with diverse poses and occlusions remains challenging. This complexity stems from intrinsic layout conflicts in the skeletal structures of multiple people, which hinder current models from accurately rendering fine details in interactive regions, leading to limb distortion and visual artifacts. In this work, we propose LAMP, a framework for pose-accurate and visually coherent multi-person image generation across various occlusion scenarios and interactions. Our key insight is to resolve spatial conflicts in overlapping skeletal regions by disentangling individual control conditions through hierarchical modeling. We introduce an occlusion-aware layered control model that leverages existing robust methods to process disentangled single-person skeletons, while designing a novel composition strategy that enables accurate perception of visibility relationships in occluded regions. Additionally, we design occlusion-aware prompt guidance to mitigate identity ambiguity among multiple individuals caused by occlusions. Both quantitative and qualitative results demonstrate LAMP’s superiority over state-of-the-art methods in generating high-quality multi-person images with diverse poses, complex occlusions, varied styles, and semantic-rich contexts.

This paper presents FlowRadar, a novel unsupervised radar detection framework that leverages Flownet2-based optical flow analysis for target detection in sea clutter environments. Unlike traditional radar detection methods that struggle with the non-linearity and non-Gaussian nature of sea clutter, FlowRadar eliminates the need for clutter filtering and labeled datasets by directly processing raw radar echoes through a spatiotemporal optical flow representation. Experimental results on Xband solid-state coherent radar data demonstrate that FlowRadar achieves a 72.8% accuracy improvement over existing Tri-feature-based methods, showcasing its potential for robust and scalable maritime surveillance. This research marks a significant advancement in radar signal processing by bridging computer vision techniques with radar physics, offering a new paradigm for target detection in complex sea environments.

Recently, several spatial-temporal memory-based methods have verified that storing intermediate frames with masks as memory helps segment target objects in videos. However, they mainly focus on better matching between the current frame and memory frames without paying attention to the quality of the memory. Consequently, frames with poor segmentation masks may be memorized, leading to error accumulation problems. Besides, the linear increase of memory frames with the growth of frame numbers limits the ability of the models to handle long videos. To this end, we propose a Quality-aware Dynamic Memory Network (QDMN) to evaluate the segmentation quality of each frame, allowing the memory bank to selectively store accurately segmented frames and prevent error accumulation. Then, we combine the segmentation quality with temporal consistency to dynamically update the memory bank and make the models can handle videos of arbitrary length. The above operation ensures the reliability of memory frames and improves the quality of memory at the frame level. Moreover, we observe that the memory features extracted from reliable frames still contain noise and have limited representation capabilities. To address this problem, we propose to perform memory enhancement and anchoring on the basis of QDMN to improve the quality of memory from the feature level, resulting in a more robust and effective network QDMN++. Our method achieves state-of-the-art performance on all popular benchmarks. Moreover, extensive experiments demonstrate that the proposed memory screening mechanism can be applied to any memory-based methods as generic plugins.

Recent years have witnessed the rapid development of general human action understanding. However, when applied to real-world applications such as sports analysis, most existing datasets are still unsatisfactory, because of the limitations in rich labels on multiple tasks, language instructions, high-quality 3D data, and diverse environments. In this paper, we present FLAG3D++, a large-scale benchmark for 3D fitness activity comprehension, which contains 180 K sequences of 60 activity categories with language instruction. FLAG3D++ features the following four aspects: 1) fine-grained annotations of the temporal intervals of actions in the untrimmed long sequences and how well these actions are performed, 2) detailed and professional language instruction to describe how to perform a specific activity, 3) accurate and dense 3D human pose captured from advanced MoCap system to handle the complex activity and large movement, 4) versatile video resources from a high-tech MoCap system, rendering software, and cost-effective smartphones in natural environments. In light of the specified features, we present two new practical applications as language-guided repetition action counting (L-RAC) and language-guided action quality assessment (L-AQA), which aim to take the language descriptions as references to count the repetitive times of an action and assess the quality of action respectively. Furthermore, we propose a Hierarchical Language-Guided Graph Convolutional Network (HL-GCN) model to better fuse the language information and skeleton sequences for L-RAC and L-AQA. To be specific, the HL-GCN performs cross-modal alignments by the early fusion of the linguistic feature and the hierarchical node features of the skeleton-based sequences encoded by the multiple intermediate graph convolutional layers. Extensive experiments show the superiority of our HL-GCN on both L-RAC and L-AQA, as well as the great research value of FLAG3D++ for various challenges, such as dynamic human mesh recovery and cross-domain human action recognition. Our dataset, source code, and trained models are made publicly available at FLAG3D++.

Referring segmentation is a fundamental vision-language task that aims to segment out an object from an image or video in accordance with a natural language description. One of the key challenges behind this task is leveraging the referring expression for highlighting relevant positions in the image or video frames. A paradigm for tackling this problem in both the image and the video domains is to leverage a powerful vision-language ("cross-modal") decoder to fuse features independently extracted from a vision encoder and a language encoder. Recent methods have made remarkable advances in this paradigm by exploiting Transformers as cross-modal decoders, concurrent to the Transformer's overwhelming success in many other vision-language tasks. Adopting a different approach in this work, we show that significantly better cross-modal alignments can be achieved through the early fusion of linguistic and visual features in intermediate layers of a vision Transformer encoder network. Based on the idea of conducting cross-modal feature fusion in the visual feature encoding stage, we propose a unified framework named Language-Aware Vision Transformer (LAVT), which leverages the well-proven correlation modeling power of a Transformer encoder for excavating helpful multi-modal context. This way, accurate segmentation results can be harvested with a light-weight mask predictor. One of the key components in the proposed system is a dense attention mechanism for collecting pixel-specific linguistic cues. When dealing with video inputs, we present the video LAVT framework and design a 3D version of this component by introducing multi-scale convolutional operators arranged in a parallel fashion, which can exploit spatio-temporal dependencies at different granularity levels. We further introduce unified LAVT as a unified framework that could handle both image and video inputs with enhanced segmentation capability on unified referring segmentation task. Our methods surpass previous state-of-the-art methods on seven benchmarks for referring image segmentation and referring video segmentation. The code to reproduce our experiments is available at LAVT-RS.

Text-to-3D generation enables the creation of 3D content with infinite possibilities. Existing methods typically involve training 3D generative models, which suffer from poor semantic alignment due to the scarcity of paired 3D data, or optimizing a 3D representation with 2D diffusion guidance, resulting in slow inference, low diversity, and Janus problems. In this paper, we introduce InstantDreamer, a model designed for text-guided 3D-aware generation in a single forward pass without requiring paired training datasets, thereby enhancing efficiency. To accomplish this, we extend score distillation to learn a 3D-aware semantics distribution. We distill priors from diffusion models into a 3D-aware generator, amortizing the optimization time required for new prompts and eliminating the necessity of paired training data. We equip the generator with hierarchical semantics conditioning, explicitly allowing the model to perceive the correspondence between the text distribution and the 3D latent space. Our elaborate designs empower our 3D generative model with multi-view semantic consistency and feed-forward 3D generation capabilities, thus eliminating the need for score distillation-based optimization for each prompt. Both quantitative and qualitative results on the mainstream benchmarks demonstrate that our InstantDreamer generates competitive multi-view semantic consistent 3D assets compared with state-of-the-art methods. Our method outperforms previous approaches in terms of CLIP R-Precision (66.31) and FID (28.47) while also exhibiting a significant boost in generation speed.

Recent advancements in pre-trained vision-language models like CLIP, have enabled the task of open-vocabulary segmentation. CLIP demonstrates impressive zero-shot capabilities in various downstream tasks that require holistic image understanding. However, due to the image-level contrastive learning and fully global feature interaction, ViT-based CLIP struggles to capture local details, resulting in poor performance in segmentation tasks. Our analysis of ViT-based CLIP reveals that anomaly tokens emerge during the forward process, attracting disproportionate attention from normal patch tokens and thereby diminishing spatial awareness. To address this issue, we propose Self-Calibrated CLIP (SC-CLIP), a training-free method that calibrates CLIP to generate finer representations while preserving its original generalization ability-without introducing new parameters or relying on additional backbones. Specifically, we mitigate the negative impact of anomaly tokens from two complementary perspectives. First, we explicitly identify the anomaly tokens and replace them based on local context. Second, we reduce their influence on normal tokens by enhancing feature discriminability and attention correlation, leveraging the inherent semantic consistency within CLIP's mid-level features. In addition, we introduce a two-pass strategy that effectively integrates multi-level features to enrich local details under the training-free setting. Together, these strategies enhance CLIP's feature representations with improved granularity and semantic coherence. Experimental results demonstrate the effectiveness of SC-CLIP, achieving state-of-the-art results across all datasets and surpassing previous methods by 9.5%. Notably, SC-CLIP boosts the performance of vanilla CLIP ViT-L/14 by 6.8 times. Furthermore, we discuss our method's applicability to other vision-language models and tasks for a comprehensive evaluation. Our source code is available at https://github.com/SuleBai/SC-CLIP.

Occupancy prediction reconstructs 3D structures of surrounding environments. It provides detailed information for autonomous driving planning and navigation. However, most existing methods heavily rely on the LiDAR point clouds to generate occupancy ground truth, which is not available in the vision-based system. In this paper, we propose an OccNeRF method for training occupancy networks without 3D ground truth. Different from previous works which consider a bounded scene, we parameterize the reconstructed occupancy fields and reorganize the sampling strategy to align with the cameras' infinite perceptive range. The neural rendering is adopted to convert occupancy fields to multi-camera depth maps, supervised by multi-frame photometric consistency. Moreover, for semantic occupancy prediction, we design several strategies to polish the prompts and filter the outputs of a pretrained open-vocabulary 2D segmentation model. Extensive experiments for both self-supervised depth estimation and 3D occupancy prediction tasks on nuScenes and SemanticKITTI datasets demonstrate the effectiveness of our method. The code is available at https://github.com/LinShan-Bin/OccNeRF.

Input encoding has proven crucial in the success of methods based on neural radiance field. Compared to the literature on general static scene modeling, input encoding for dynamic hand modeling has been less explored. However, this aspect is critical to the modeling of deformation and rendering, as it maps a sampled point in space to the representation containing all the information associated with dynamic hand for inferring the geometry and appearance property of this point. The design of input encoding determines how well the neural network can learn for photo-realistic hand rendering. We offer an in-depth examination of this key component and introduceDEHand, a new representation utilizingDeformableEncoding for photo-realistic free-view and free-poseHandrendering. DEHand leverages deformable encoding with a latent code map to achieve high-quality, pose-controlled rendering. Deformable encoding is achieved by adapting static input encoding techniques for the view synthesis of dynamic hands, using parametric hand mesh model as a proxy to construct encodings that map sampled points into a space capable of integrating over different poses and providing rich information for hand modeling. Our findings demonstrate that with our deformable encoding, a single Multilayer Perceptron (MLP) can achieve high-quality dynamic hand rendering, learning solely from images. Extensive experiments on InterHand2.6M validate the superior rendering quality of our method and the effectiveness of each component in our design.

Weakly Supervised temporal Article Grounding (WSAG) is a challenging and practical task in video understanding. Specifically, given a video and a relevant article, whose sentences are at different semantic scales, WSAG aims to localize corresponding video segments for all “groundable” sentences. Compared to other grounding tasks, e.g., localizing one target segment with respect to a given sentence query, WSAG confronts an essential obstacle rooted in the intricate multi-scale information inherent within both textual and visual modalities. Existing methods overlook the modeling and alignment of such structured information present in multi-scale video segments and hierarchical textual content. To this end, we propose a Multi-Scale Video-Text Correspondence Learning (MVTCL) framework, which enhances the grounding performance in complex scenes by modeling multi-scale semantic correspondence both within and between modalities. Specifically, MVTCL initially aggregates video content spanning distinct temporal scales and leverages hierarchical textual relationships in both temporal and semantic dimensions via a semantic calibration module. Then multi-scale contrastive learning module is introduced to generate more discriminative representations by selecting typical contexts and performing inter-video contrastive learning. Through the multi-scale semantic calibration architecture and supervision design, our method achieves new state-of-the-art performance on existing WSAG benchmarks.

This paper studies the spatio-temporal video grounding task, which aims to localize a spatio-temporal tube in an untrimmed video based on the given text description of an event. Existing one-stage approaches suffer from insufficient space-time interaction in two aspects: i) less precise prediction of event temporal boundaries, and ii) inconsistency in object prediction for the same event across adjacent frames. To address these issues, we propose a framework of Comprehensive Space-Time entAnglement (CoSTA) to densely entangle space-time multi-modal features for spatio-temporal localization. Specifically, we propose a space-time collaborative encoder to extract comprehensive video features and leverage Transformer to perform spatio-temporal multi-modal understanding. Our entangled decoder couples temporal boundary prediction and spatial localization via an entangled query, boasting an enhanced ability to capture object-event relationships. We conduct extensive experiments on the challenging benchmarks of HC-STVG and VidSTG, where CoSTA outperforms existing state-of-the-art methods, demonstrating its effectiveness for this task.

Vision-Language Transformers (VLTs) have shown great success recently, but are meanwhile accompanied by heavy computation costs, where a major reason can be attributed to the large number of visual and language tokens. Existing token pruning research for compressing VLTs mainly follows a single-modality-based scheme yet ignores the critical role of aligning different modalities for guiding the token pruning process, causing the important tokens for one modality to be falsely pruned in another modality branch. Meanwhile, existing VLT pruning works also lack the flexibility to dynamically compress each layer based on different input samples. To this end, we propose a novel framework named Multimodal Alignment-Guided Dynamic Token Pruning (MADTP) for accelerating various VLTs. Specifically, we first introduce a well-designed Multi-modality Alignment Guidance (MAG) module that can align features of the same semantic concept from different modalities, to ensure the pruned tokens are less important for all modalities. We further design a novel Dynamic Token Pruning (DTP) module, which can adaptively adjust the token compression ratio in each layer based on different input instances. Extensive experiments on various benchmarks demonstrate that MADTP significantly reduces the computational complexity of kinds of multimodal models while preserving competitive performance. Notably, when applied to the BLIP model in the NLVR2 dataset, MADTP can reduce the GFLOPs by 80% with less than 4% performance degradation. The code is available at https://github.com/double125IMADTP.

We propose a method to efficiently equip the Segment Anything Model (SAM) with the ability to generate regional captions. SAM presents strong generalizability to segment anything while is short for semantic understanding. By introducing a lightweight query-based feature mixer, we align the region-specific features with the embedding space of language models for later caption generation. As the number of trainable parameters is small (typically in the order of tens of millions), it costs less computation, less memory usage, and less communication bandwidth, resulting in both fast and scalable training. To address the scarcity problem of regional caption data, we propose to first pretrain our model on objection detection and segmentation tasks. We call this step weak supervision pretraining since the pretraining data only contains category names instead of full-sentence descriptions. The weak supervision pretraining al-lows us to leverage many publicly available object detection and segmentation datasets. We conduct extensive experiments to demonstrate the superiority of our method and validate each design choice. This work serves as a step-ping stone towards scaling up regional captioning data and sheds light on exploring efficient ways to augment SAM with regional semantics. The project page, along with the associated code, can be accessed via the following link.

This paper studies open-vocabulary segmentation (OVS) through calibrating in-vocabulary and domain-biased embedding space with generalized contextual prior of CLIP. As the core of open-vocabulary understanding, alignment of visual content with the semantics of unbounded text has become the bottleneck of this field. To address this challenge, recent works propose to utilize CLIP as an additional classifier and aggregate model predictions with CLIP classification results. Despite their remarkable progress, performance of OVS methods in relevant scenarios is still unsatisfactory compared with supervised counterparts. We attribute this to the in-vocabulary embedding and domain-biased CLIP prediction. To this end, we present a Semantic-assisted CAlibration Network (SCAN). In SCAN, we incor-porate generalized semantic prior of CLIP into proposal embedding to avoid collapsing on known categories. Besides, a contextual shift strategy is applied to mitigate the lack of global context and unnatural background noise. With above designs, SCAN achieves state-of-the-art performance on all popular open-vocabulary segmentation benchmarks. Furthermore, we also focus on the problem of existing evaluation system that ignores semantic duplication across categories, and propose a new metric called Semantic-Guided IoU (SG-IoU). Code is available here.

This paper aims to achieve universal segmentation of arbitrary semantic level. Despite significant progress in recent years, specialist segmentation approaches are limited to specific tasks and data distribution. Retraining a new model for adaptation to new scenarios or settings takes expensive computation and time cost, which raises the demand for versatile and universal segmentation model that can cater to various granularity. Although some attempts have been made for unifying different segmentation tasks or generalization to various scenarios, limitations in the definition of paradigms and input-output spaces make it difficult for them to achieve accurate understanding of content at arbitrary granularity. To this end, we present UniLSeg, a universal segmentation model that can perform segmentation at any semantic level with the guidance of language instructions. For training UniLSeg, we reorganize a group of tasks from original diverse distributions into a unified data format, where images with texts describing segmentation targets as input and corresponding masks are output. Combined with a automatic annotation engine for utilizing numerous unlabeled data, UniLSeg achieves excellent performance on various tasks and settings, surpassing both specialist and unified segmentation models. Code is available here.

In this paper, we propose an accurate post-training quantization framework of diffusion models (APQ-DM) for efficient image generation. Conventional quantization frameworks learn shared quantization functions for tensor discretization regardless of the generation timesteps in diffusion models, while the activation distribution differs significantly across various timesteps. Meanwhile, the calibration images are acquired in random timesteps which fail to provide sufficient information for generalizable quantization function learning. Both issues cause sizable quantization errors with obvious image generation performance degradation. On the contrary, we design distribution-aware quantization functions for activation discretization in different timesteps and search the optimal timesteps for informative calibration image generation, so that our quantized diffusion model can reduce the discretization errors with negligible computational overhead. Specifically, we partition various timestep quantization functions into different groups according to the importance weights, which are optimized by differentiable search algorithms. We also extend structural risk minimization principle for informative calibration image generation to enhance the generalization ability in the deployment of quantized diffusion model. Extensive experimental results show that our method outperforms the state-of-the-art post-training quantization of diffusion model by a sizable margin with similar computational cost11Code is available at https://github.com/ChangyuanWang17/APQ-DM.

Recent Vision Transformer Compression (VTC) works mainly follow a two-stage scheme, where the importance score of each model unit is first evaluated or preset in each submodule, followed by the sparsity score evaluation ac-cording to the target sparsity constraint. Such a separate evaluation process induces the gap between importance and sparsity score distributions, thus causing high search costs for VTC. In this work, for the first time, we investigate how to integrate the evaluations of importance and sparsity scores into a single stage, searching the optimal subnets in an effi-cient manner. Specifically, we present OFB, a cost-efficient approach that simultaneously evaluates both importance and sparsity scores, termed Once for Both (OFB), for VTC. First, a bi-mask scheme is developed by entangling the importance score and the differentiable sparsity score to jointly deter-mine the pruning potential (prunability) of each unit. Such a bi-mask search strategy is further used together with a proposed adaptive one-hot loss to realize the progressive-and-efficient search for the most important subnet. Finally, Progressive Masked Image Modeling (PMIM) is proposed to regularize the feature space to be more representative during the search process, which may be degraded by the dimension reduction. Extensive experiments demonstrate that OFB can achieve superior compression performance over state-of-the-art searching-based and pruning-based methods under various Vision Transformer architectures, meanwhile pro-moting search efficiency significantly, e.g., costing one GPU search day for the compression of DeiT-S on ImageNet-1K.

Video quality assessment (VQA) is a challenging problem due to the numerous factors that can affect the perceptual quality of a video, e.g., content attractiveness, distortion type, motion pattern, and level. However, annotating the Mean opinion score (MOS) for videos is expensive and time-consuming, which limits the scale of VQA datasets, and poses a significant obstacle for deep learning-based methods. In this paper, we propose a VQA method named PTM-VQA, which leverages PreTrained Models to transfer knowledge from models pretrained on various pre-tasks, enabling benefits for VQA from different aspects. Specifically, we extract features of videos from different pretrained models with frozen weights and integrate them to generate representation. Since these models possess var-ious fields of knowledge and are often trained with labels irrelevant to quality, we propose an Intra-Consistency and Inter-Divisibility (ICID) loss to impose constraints on features extracted by multiple pretrained models. The intra-consistency constraint ensures that features extracted by different pretrained models are in the same unified quality-aware latent space, while the inter-divisibility introduces pseudo clusters based on the annotation of samples and tries to separate features of samples from different clusters. Furthermore, with a constantly growing number of pretrained models, it is crucial to determine which models to use and how to use them. To address this problem, we propose an efficient scheme to select suitable candidates. Models with better clustering performance on VQA datasets are chosen to be our candidates. Extensive experiments demonstrate the effectiveness of the proposed method.

In this paper, we investigate a new problem called narrative action evaluation (NAE). NAE aims to generate professional commentary that evaluates the execution of an action. Unlike traditional tasks such as score-based action qual-ity assessment and video captioning involving superficial sentences, NAE focuses on creating detailed narratives in natural language. These narratives provide intricate descriptions of actions along with objective evaluations. NAE is a more challenging task because it requires both narrative flex-ibility and evaluation rigor. One existing possible solution is to use multi-task learning, where narrative language and evaluative information are predicted separately. However, this approach results in reduced performance for individual tasks because of variations between tasks and differences in modality between language information and evaluation information. To address this, we propose a prompt-guided multimodal interaction framework. This framework utilizes a pair of transformers to facilitate the interaction between different modalities of information. It also uses prompts to transform the score regression task into a video-text matching task, thus enabling task interactivity. To support further research in this field, we re-annotate the MTL-AQA and FineGym datasets with high-quality and comprehensive action narration. Additionally, we establish benchmarks for NAE. Extensive experiment results prove that our method outperforms separate learning methods and naive multi-task learning methods. Data and code are released at here.

The recovery of occluded human meshes presents challenges for current methods due to the difficulty in extracting effective image features under severe occlusion. In this paper, we introduce DPMesh, an innovative framework for occluded human mesh recovery that capitalizes on the pro-found diffusion prior about object structure and spatial relationships embedded in a pre-trained text-to-image diffusion model. Unlike previous methods reliant on conventional backbones for vanilla feature extraction, DPMesh seamlessly integrates the pre-trained denoising U-Net with potent knowledge as its image backbone and performs a single-step inference to provide occlusion-aware information. To enhance the perception capability for occluded poses, DPMesh incorporates well-designed guidance via condition injection, which produces effective controls from 2D observations for the denoising U-Net. Furthermore, we explore a dedicated noisy key-point reasoning approach to mitigate disturbances arising from occlusion and crowded scenarios. This strategy fully unleashes the perceptual capability of the diffusion prior, thereby enhancing accuracy. Extensive experiments affirm the efficacy of our frame-work, as we outperform state-of-the-art methods on both occlusion-specific and standard datasets. The persuasive results underscore its ability to achieve precise and robust 3D human mesh recovery, particularly in challenging scenarios involving occlusion and crowded scenes. Code is available at https://github.com/EternalEvan/DPMesh.

Image enhancement holds extensive applications in real-world scenarios due to complex environments and limitations of imaging devices. Conventional methods are often constrained by their tailored models, resulting in diminished robustness when confronted with challenging degradation conditions. In response, we propose FlowIE, a simple yet highly effective flow-based image enhancement framework that estimates straight-line paths from an elementary distribution to high-quality images. Unlike previous diffusion-based methods that suffer from long-time inference, FlowIE constructs a linear many-to-one transport mapping via conditioned rectified flow. The rectification straightens the trajectories of probability transfer, accelerating inference by an order of magnitude. This design enables our FlowIE to fully exploit rich knowledge in the pretrained diffusion model, rendering it well-suited for various real-world applications. Moreover, we devise a faster inference algorithm, inspired by Lagrange's Mean Value Theorem, harnessing midpoint tangent direction to optimize path estimation, ultimately yielding visually superior results. Thanks to these designs, our FlowIE adeptly manages a diverse range of enhancement tasks within a concise sequence of fewer than 5 steps. Our contributions are rigorously validated through comprehensive experiments on synthetic and real-world datasets, unveiling the compelling efficacy and efficiency of our proposed FlowIE. Code is available at https://github.com/EternalEvan/FlowIE.

Composing simple actions into complex actions is crucial yet challenging. Existing methods largely rely on language annotations to discern composable latent semantics, which is costly and labor-intensive. In this study, we introduce a novel framework to generate compositional actions without language auxiliaries. Our approach consists of three components: Action Coupling, Conditional Action Generation, and Decoupling Refinement. Action Coupling integrates two sub-actions to generate pseudo-training examples. Then, a conditional generative model, CVAE is employed to facilitate the diverse generation. Decoupling Refinement leverages a self-supervised pre-trained model MAE to ensure semantic consistency between sub-actions and compositional actions. Due to the lack of existing datasets containing both sub-actions and compositional actions, we create two new datasets, named HumanAct-C and UESTC-C. Both qualitative and quantitative assessments are conducted to show our efficacy.1

Recent work demonstrates that, post-training large language models with open-domain instruction following data have achieved colossal success. Simultaneously, human Chatbot Arena has emerged as one of the most reasonable benchmarks for model evaluation and developmental guidance. However, the processes of manually curating high-quality training data and utilizing online human evaluation platforms are both expensive and limited. To mitigate the manual and temporal costs associated with post-training, this paper introduces a Simulated Chatbot Arena named WizardArena, which is fully based on and powered by open-source LLMs. For evaluation scenario, WizardArena can efficiently predict accurate performance rankings among different models based on offline test set. For training scenario, we simulate arena battles among various state-of-the-art models on a large scale of instruction data, subsequently leveraging the battle results to constantly enhance target model in both the supervised fine-tuning and reinforcement learning . Experimental results demonstrate that our WizardArena aligns closely with the online human arena rankings, and our models trained on offline extensive battle data exhibit significant performance improvements during SFT, DPO, and PPO stages.

In this paper, we propose a post-training quantization framework of large vision-language models (LVLMs) for efficient multi-modal inference. Conventional quantization methods sequentially search the layer-wise rounding functions by minimizing activation discretization errors, which fails to acquire optimal quantization strategy without considering cross-layer dependency. On the contrary, we mine the cross-layer dependency that significantly influences discretization errors of the entire vision-language model, and embed this dependency into optimal quantization strategy searching with low search cost. Specifically, we observe the strong correlation between the activation entropy and the cross-layer dependency concerning output discretization errors. Therefore, we employ the entropy as the proxy to partition blocks optimally, which aims to achieve satisfying trade-offs between discretization errors and the search cost. Moreover, we optimize the visual encoder to disentangle the cross-layer dependency for fine-grained decomposition of search space, so that the search cost is further reduced without harming the quantization accuracy. Experimental results demonstrate that our method compresses the memory by 2.78x and increase generate speed by 1.44x about 13B LLaVA model without performance degradation on diverse multi-modal reasoning tasks.

We introduce a radiance representation that is both structured and fully explicit and thus greatly facilitates 3D generative modeling. Existing radiance representations either require an implicit feature decoder, which significantly degrades the modeling power of the representation, or are spatially unstructured, making them difficult to integrate with mainstream 3D diffusion methods. We derive GaussianCube by first using a novel densification-constrained Gaussian fitting algorithm, which yields high-accuracy fitting using a fixed number of free Gaussians, and then rearranging these Gaussians into a predefined voxel grid via Optimal Transport. Since GaussianCube is a structured grid representation, it allows us to use standard 3D U-Net as our backbone in diffusion modeling without elaborate designs. More importantly, the high-accuracy fitting of the Gaussians allows us to achieve a high-quality representation with orders of magnitude fewer parameters than previous structured representations for comparable quality, ranging from one to two orders of magnitude. The compactness of GaussianCube greatly eases the difficulty of 3D generative modeling. Extensive experiments conducted on unconditional and class-conditioned object generation, digital avatar creation, and text-to-3D synthesis all show that our model achieves state-of-the-art generation results both qualitatively and quantitatively, underscoring the potential of GaussianCube as a highly accurate and versatile radiance representation for 3D generative modeling.

In this work, we introduce the Geometry-Aware Large Reconstruction Model (GeoLRM), an approach which can predict high-quality assets with 512k Gaussians and 21 input images in only 11 GB GPU memory. Previous works neglect the inherent sparsity of 3D structure and do not utilize explicit geometric relationships between 3D and 2D images. This limits these methods to a low-resolution representation and makes it difficult to scale up to the dense views for better quality. GeoLRM tackles these issues by incorporating a novel 3D-aware transformer structure that directly processes 3D points and uses deformable cross-attention mechanisms to effectively integrate image features into 3D representations. We implement this solution through a two-stage pipeline: initially, a lightweight proposal network generates a sparse set of 3D anchor points from the posed image inputs; subsequently, a specialized reconstruction transformer refines the geometry and retrieves textural details. Extensive experimental results demonstrate that GeoLRM significantly outperforms existing models, especially for dense view inputs. We also demonstrate the practical applicability of our model with 3D generation tasks, showcasing its versatility and potential for broader adoption in real-world applications. The project page: https://linshan-bin.github.io/GeoLRM/.

This paper focuses on the Referring Image Segmentation (RIS) task, which aims to segment objects from an image based on a given language description, having significant potential in practical applications such as food safety detection. Recent advances using the attention mechanism for cross-modal interaction have achieved excellent progress. However, current methods tend to lack explicit principles of interaction design as guidelines, leading to inadequate cross-modal comprehension. Additionally, most previous works use a single-modal mask decoder for prediction, losing the advantage of full cross-modal alignment. To address these challenges, we present a Fully Aligned Network (FAN) that follows four cross-modal interaction principles. Under the guidance of reasonable rules, our FAN achieves state-of-the-art performance on the prevalent RIS benchmarks (RefCOCO, RefCOCO+, G-Ref) with a simple architecture.

Repetitive action counting (RAC) aims to estimate the number of class-agnostic action occurrences in a video without exemplars. Most current RAC methods rely on a raw frame-to-frame similarity representation for period prediction. However, this approach can be significantly disrupted by common noise such as action interruptions and inconsistencies, leading to sub-optimal counting performance in realistic scenarios. In this paper, we introduce a foreground localization optimization objective into similarity representation learning to obtain more robust and efficient video features. We propose a Localization-Aware Multi-Scale Representation Learning (LMRL) framework. Specifically, we apply a Multi-Scale Period-Aware Representation (MPR) with a scale-specific design to accommodate various action frequencies and learn more flexible temporal correlations. Furthermore, we introduce the Repetition Foreground Localization (RFL) method, which enhances the representation by coarsely identifying periodic actions and incorporating global semantic information. These two modules can be jointly optimized, resulting in a more discerning periodic action representation. Our approach significantly reduces the impact of noise, thereby improving counting accuracy. Additionally, the framework is designed to be scalable and adaptable to different types of video content. Experimental results on the RepCountA and UCFRep datasets demonstrate that our proposed method effectively handles repetitive action counting.

This study introduces a novel approach for the estimation of seismic source parameters using a multi-task learning network that incorporates a Fourier Transformer architecture. The Fourier Transformer is designed to extract information from both the time and frequency domains, which reduces the time complexity by utilizing Fast Fourier Transform (FFT) in place of the traditional attention mechanism in the Transformer encoder. The network consists of a shared encoder for general feature extraction and four task-specific decoders for parameter estimation. The model is both lightweight and accurate, capable of simultaneously estimating magnitude, epicentral distance, p travel time, and depth based on a 30-second single-station waveform. The proposed approach was validated using the Stanford Earthquake dataset (STEAD) and compared with the state-of-the-art techniques. The results show standard deviations of 0.19 for magnitude, 3.77 km for epicentral distance, 0.46 s for p travel time, and 5.77 km for depth, with a lower error and a faster response compared to the existing prediction framework. Code is available at https://github.com/KG-TSI-Civil/MFTnet.

Person-agnostic face swapping has gained significant attention in recent years, as it offers the potential to enhance various real-world applications by combining high fidelity and identity consistency. However, conventional face swapping methods often rely on intricate adjustments of different loss functions, leading to instability during both the training and inference stages. In this work, we propose a simple yet effective framework namedStableSwapwith a reversible autoencoder to modify the face in a shared latent space. Our approach capitalizes on the information-rich image latent codes to tackle the challenges of complex editing tasks, utilizing the abundant details present in both the source and target faces. To ensure an expressive and robust latent space, we employ a latent alignment approach with perceptual and adversarial losses to optimize the autoencoder. Additionally, we devise a multi-stage identity injection module that samples multiple features with different facial priors and incorporates them to guide the latent image manipulation. By leveraging attention-based blocks, we fuse these futures and update the latent code in a mask-conditioned manner. Both quantitative and qualitative results on the mainstream benchmarks demonstrate that our StableSwap generates competitive identity-consistent swapped faces compared with state-of-the-art methods. Our method outperforms previous approaches in terms of ID Retrieval (98.68) and FID (2.49), while also exhibiting enhanced stability during model training. Beyond this, our model achieves region-controllable face swapping with the capability to perform more fine-grained operations in latent space.

Referring image segmentation segments an image from a language expression. With the aim of producing high-quality masks, existing methods often adopt iterative learning approaches that rely on RNNs or stacked attention layers to refine vision-language features. Despite their complexity, RNN-based methods are subject to specific encoder choices, while attention-based methods offer limited gains. In this work, we introduce a simple yet effective alternative for progressively learning discriminative multi-modal features. The core idea of our approach is to leverage a continuously updated query as the representation of the target object and at each iteration, strengthen multi-modal features strongly correlated to the query while weakening less related ones. As the query is initialized by language features and successively updated by object features, our algorithm gradually shifts from being localization-centric to segmentation-centric. This strategy enables the incremental recovery of missing object parts and/or removal of extraneous parts through iteration. Compared to its counterparts, our method is more versatile—it can be plugged into prior arts straightforwardly and consistently bring improvements. Experimental results on the challenging datasets of RefCOCO, RefCOCO+, and G-Ref demonstrate its advantage with respect to the state-of-the-art methods.

With the continuously thriving popularity around the world, fitness activity analytic has become an emerging research topic in computer vision. While a variety of new tasks and algorithms have been proposed recently, there are growing hunger for data resources involved in high-quality data, fine-grained labels, and diverse environments. In this paper, we present FLAG3D, a large-scale 3D fitness activity dataset with language instruction containing 180K sequences of 60 categories. FLAG3D features the following three aspects: 1) accurate and dense 3D human pose captured from advanced MoCap system to handle the complex activity and large movement, 2) detailed and professional language instruction to describe how to perform a specific activity, 3) versatile video resources from a high-tech MoCap system, rendering software, and cost-effective smartphones in natural environments. Extensive experiments and in-depth analysis show that FLAG3D contributes great research value for various challenges, such as cross-domain human action recognition, dynamic human mesh recovery, and language-guided human action generation. Our dataset and source code are publicly available at https://andytang15.github.io/FLAG3D.

Action quality assessment (AQA) has become an emerging topic since it can be extensively applied in numerous scenarios. However, most existing methods and datasets focus on single-person short-sequence scenes, hindering the application of AQA in more complex situations. To address this issue, we construct a new multi-person long-form video dataset for action quality assessment named LOGO. Distinguished in scenario complexity, our dataset contains 200 videos from 26 artistic swimming events with 8 athletes in each sample along with an average duration of 204.2 seconds. As for richness in annotations, LOGO includes formation labels to depict group information of multiple athletes and detailed annotations on action procedures. Furthermore, we propose a simple yet effective method to model relations among athletes and reason about the potential temporal logic in long-form videos. Specifically, we design a group-aware attention module, which can be easily plugged into existing AQA methods, to enrich the clip-wise representations based on contextual group information. To benchmark LOGO, we systematically conduct investigations on the performance of several popular methods in AQA and action segmentation. The results reveal the challenges our dataset brings. Extensive experiments also show that our approach achieves state-of-the-art on the LOGO dataset. The dataset and code will be released at https://github.com/shiyi-zh0408/LOGO.

Video-language pre-trained models have shown remarkable success in guiding video question-answering (VideoQA) tasks. However, due to the length of video sequences, training large-scale video-based models incurs considerably higher costs than training image-based ones. This motivates us to leverage the knowledge from image-based pretraining, despite the obvious gaps between image and video domains. To bridge these gaps, in this paper, we propose Tem-adapter, which enables the learning of temporal dynamics and complex semantics by a visual Temporal Aligner and a textual Semantic Aligner. Unlike conventional pretrained knowledge adaptation methods that only concentrate on the downstream task objective, the Temporal Aligner introduces an extra language-guided autoregressive task aimed at facilitating the learning of temporal dependencies, with the objective of predicting future states based on historical clues and language guidance that describes event progression. Besides, to reduce the semantic gap and adapt the textual representation for better event description, we introduce a Semantic Aligner that first designs a template to fuse question and answer pairs as event descriptions and then learns a Transformer decoder with the whole video sequence as guidance for refinement. We evaluate Tem-adapter and different pre-train transferring methods on two VideoQA benchmarks, and the significant performance improvement demonstrates the effectiveness of our method.1

Recent advancements in pre-trained vision-language models, such as CLIP, have enabled the segmentation of arbitrary concepts solely from textual inputs, a process commonly referred to as open-vocabulary semantic segmentation (OVS). However, existing OVS techniques confront a fundamental challenge: the trained classifier tends to over-fit on the base classes observed during training, resulting in suboptimal generalization performance to unseen classes. To mitigate this issue, recent studies have proposed the use of an additional frozen pre-trained CLIP for classification. Nonetheless, this approach incurs heavy computational overheads as the CLIP vision encoder must be repeatedly forward-passed for each mask, rendering it impractical for real-world applications. To address this challenge, our objective is to develop a fast OVS model that can perform comparably or better without the extra computational burden of the CLIP image encoder during inference. To this end, we propose a core idea of preserving the generalizable representation when fine-tuning on known classes. Specifically, we introduce a text diversification strategy that generates a set of synonyms for each training category, which prevents the learned representation from collapsing onto specific known category names. Additionally, we employ a text-guided knowledge distillation method to preserve the generalizable knowledge of CLIP. Extensive experiments demonstrate that our proposed model achieves robust generalization performance across various datasets. Furthermore, we perform a preliminary exploration of open-vocabulary video segmentation and present a benchmark that can facilitate future open-vocabulary research in the video domain.

In this paper, we propose Skip-Plan, a condensed action space learning method for procedure planning in instructional videos. Current procedure planning methods all stick to the state-action pair prediction at every timestep and generate actions adjacently. Although it coincides with human intuition, such a methodology consistently struggles with high-dimensional state supervision and error accumulation on action sequences. In this work, we abstract the procedure planning problem as a mathematical chain model. By skipping uncertain nodes and edges in action chains, we transfer long and complex sequence functions into short but reliable ones in two ways. First, we skip all the intermediate state supervision and only focus on action predictions. Second, we decompose relatively long chains into multiple short sub-chains by skipping unreliable intermediate actions. By this means, our model explores all sorts of reliable sub-relations within an action sequence in the condensed action space. Extensive experiments show Skip-Plan achieves state-of-the-art performance on the CrossTask and COIN benchmarks for procedure planning.

Generating full-body and multi-genre dance sequences from given music is a challenging task, due to the limitations of existing datasets and the inherent complexity of the fine-grained hand motion and dance genres. To address these problems, we propose FineDance, which contains 14.6 hours of music-dance paired data, with fine-grained hand motions, fine-grained genres (22 dance genres), and accurate posture. To the best of our knowledge, FineDance is the largest music-dance paired dataset with the most dance genres. Additionally, to address monotonous and unnatural hand movements existing in previous methods, we propose a full-body dance generation network, which utilizes the diverse generation capabilities of the diffusion model to solve monotonous problems, and use expert nets to solve unreal problems. To further enhance the genre-matching and long-term stability of generated dances, we propose a Genre&Coherent aware Retrieval Module. Besides, we propose a novel metric named Genre Matching Score to evaluate the genre-matching degree between dance and music. Quantitative and qualitative experiments demonstrate the quality of FineDance, and the state-of-the-art performance of FineNet. The FineDance Dataset and more qualitative samples can be found at website.

In this paper, we propose a context-aware inpainter-refiner (CAIR) framework for skeleton-based human motion completion, which aims to restore the damaged area of a given corrupted sequence in spatio-temporal domain. Conventional methods usually utilize convolutional neural networks to inpaint the corrupted sequence treated as a image. In comparison, we devise a two-stage "inpainting-refining" framework and further design a context-aware graph convolutional module (CA-GCM), which employs graph convolutional operation and mask-aware residual connection to update values of corrupted area. In this way, our CAIR is capable of mapping and associating intact and corrupted joints implicitly, and thus optimizes the result generated by our inpainting network effectively. Extensive experimental results on NTU60 and Human3.6M datasets demonstrate that our method outperforms our own baseline and previous methods.

In this paper, we propose an HOI-aware adaptive network named AdaAct for weakly-supervised action segmentation. Most existing methods learn a fixed network to predict the action of each frame with the neighboring frames. However, this would result in ambiguity when estimating similar actions, such as pouring juice and pouring coffee. To address this, we aim to exploit temporally global but spatially local human-object interactions (HOI) as video-level prior knowledge for action segmentation. The long-term HOI sequence provides crucial contextual information to distinguish ambiguous actions, where our network dynamically adapts to the given HOI sequence at test time. More specifically, we first design a video HOI encoder that extracts, selects, and integrates the most representative HOI throughout the video. Then, we propose a two-branch HyperNetwork to learn an adaptive temporal encoder, which automatically adjusts the parameters based on the HOI information of various videos on the fly. Extensive experiments on two widely-used datasets including Breakfast and 50Salads demonstrate the effectiveness of our method under different evaluation metrics.

Referring expression comprehension aims to localize a natural language description in an image. Using location priors to help reduce inaccuracies in cross-modal alignments is the state of the art for CNN-based methods tackling this problem. Recent Transformer-based models cast aside this idea, making the case for steering away from hand-designed components. In this work, we propose LUNA, which uses language as continuing anchors to guide box prediction in a Transformer decoder, and thus show that language-guided location priors can be effectively exploited in a Transformer-based architecture. Our method first initializes an anchor box from the input expression via a small "proto-decoder,'' and then uses this anchor and its refined successors as location guidance in a modified Transformer decoder. At each decoder layer, the anchor box is first used as a query for gathering multi-modal context, and then updated based on the gathered context (producing the next, refined anchor). In the end, a lightweight assessment pathway evaluates the quality of all produced anchors, yielding the final prediction in a dynamic way. This approach allows box decoding to be conditioned on learned anchors, which facilitates accurate grounding, as we shown in the experiments. Our method outperforms existing state-of-the-art methods on the datasets of ReferIt Game, RefCOCO/+/g, and Flickr30K Entities.

Video quality assessment (VQA) has attracted growing attention in recent years. While the great expense of annotating large-scale VQA datasets has become the main obstacle for current deep-learning methods. To surmount the constraint of insufficient training data, in this paper, we first consider the complete range of video distribution diversity (i.e. content, distortion, motion) and employ diverse pretrained models (e.g. architecture, pretext task, pre-training dataset) to benefit quality representation. An Adaptive Diverse Quality-aware feature Acquisition (Ada-DQA) framework is proposed to capture desired quality-related features generated by these frozen pretrained models. By leveraging the Quality-aware Acquisition Module (QAM), the framework is able to extract more essential and relevant features to represent quality. Finally, the learned quality representation is utilized as supplementary supervisory information, along with the supervision of the labeled quality score, to guide the training of a relatively lightweight VQA model in a knowledge distillation manner, which largely reduces the computational cost during inference. Experimental results on three mainstream no-reference VQA benchmarks clearly show the superior performance of Ada-DQA in comparison with current state-of-the-art approaches without using extra training data of VQA.

Due to the high price and heavy energy consumption of GPUs, deploying deep models on IoT devices such as microcontrollers makes significant contributions for ecological AI. Conventional methods successfully enable convolutional neural network inference of high resolution images on microcontrollers, while the framework for vision transformers that achieve the state-of-the-art performance in many vision applications still remains unexplored. In this paper, we propose a hardware-algorithm co-optimizations method called MCUFormer to deploy vision transformers on microcontrollers with extremely limited memory, where we jointly design transformer architecture and construct the inference operator library to fit the memory resource constraint. More specifically, we generalize the one-shot network architecture search (NAS) to discover the optimal architecture with highest task performance given the memory budget from the microcontrollers, where we enlarge the existing search space of vision transformers by considering the low-rank decomposition dimensions and patch resolution for memory reduction. For the construction of the inference operator library of vision transformers, we schedule the memory buffer during inference through operator integration, patch embedding decomposition, and token overwriting, allowing the memory buffer to be fully utilized to adapt to the forward pass of the vision transformer. Experimental results demonstrate that our MCUFormer achieves 73.62\% top-1 accuracy on ImageNet for image classification with 320KB memory on STM32F746 microcontroller. Code is available at https://github.com/liangyn22/MCUFormer.

This paper studies referring video object segmentation (RVOS) by boosting video-level visual-linguistic alignment. Recent approaches model the RVOS task as a sequence prediction problem and perform multi-modal interaction as well as segmentation for each frame separately. However, the lack of a global view of video content leads to difficulties in effectively utilizing inter-frame relationships and understanding textual descriptions of object temporal variations. To address this issue, we propose Semantic-assisted Object Cluster (SOC), which aggregates video content and textual guidance for unified temporal modeling and cross-modal alignment. By associating a group of frame-level object embeddings with language tokens, SOC facilitates joint space learning across modalities and time steps. Moreover, we present multi-modal contrastive supervision to help construct well-aligned joint space at the video level. We conduct extensive experiments on popular RVOS benchmarks, and our method outperforms state-of-the-art competitors on all benchmarks by a remarkable margin. Besides, the emphasis on temporal coherence enhances the segmentation stability and adaptability of our method in processing text expressions with temporal variations. Code is available at https://github.com/RobertLuo1/NeurIPS2023_SOC.

Many video understanding tasks require analyzing multishot videos, but existing datasets for video object segmentation (VOS) only consider single-shot videos. To address this challenge, we collected a new dataset-YouMVaS-of 200 popular YouTube videos spanning ten genres, where each video is on average five minutes long and with 75 shots. We selected recurring actors and annotated 431K segmentation masks at a frame rate of six, exceeding previous datasets in average video duration, object variation, and narrative structure complexity. We incorporated good practices of model architecture design, memory management, and multi-shot tracking into an existing video segmentation method to build competitive baseline methods. Through error analysis, we found that these baselines still fail to cope with cross-shot appearance variation on our YouMVOS dataset. Thus, our dataset poses new challenges in multi-shot segmentation towards better video analysis. Data, code, and pre-trained models are available at https://donglaiw.github.io/proj/youMVOS

Referring image segmentation is a fundamental vision-language task that aims to segment out an object referred to by a natural language expression from an image. One of the key challenges behind this task is leveraging the referring expression for highlighting relevant positions in the image. A paradigm for tackling this problem is to leverage a powerful vision-language (“cross-madal”) decoder to fuse features independently extracted from a vision encoder and a language encoder. Recent methods have made remarkable advancements in this paradigm by exploiting Transformers as cross-modal decoders, concurrent to the Transformer's overwhelming success in many other vision-language tasks. Adopting a different approach in this work, we show that significantly better cross-modal alignments can be achieved through the early fusion of linguistic and visual features in intermediate layers of a vision Transformer encoder network. By conducting cross-modal feature fusion in the visual feature encoding stage, we can leverage the well-proven correlation modeling power of a Transformer encoder for excavating helpful multi-modal context. This way, accurate segmentation results are readily harvested with a light-weight mask predictor. Without bells and whistles, our method surpasses the previous state-of-the-art methods on Ref CoCo, RefCOCO+, and G-Ref by large margins.

Dense 3D reconstruction from a stream of depth images is the key to many mixed reality and robotic applications. Although methods based on Truncated Signed Distance Function (TSDF) Fusion have advanced the field over the years, the TSDF volume representation is confronted with striking a balance between the robustness to noisy measurements and maintaining the level of detail. We present Bi-level Neural Volume Fusion (BNV-Fusion), which leverages recent advances in neural implicit representations and neural rendering for dense 3D reconstruction. In order to incrementally integrate new depth maps into a global neural implicit representation, we propose a novel bi-level fusion strategy that considers both efficiency and reconstruction quality by design. We evaluate the proposed method on multiple datasets quantitatively and qualitatively, demonstrating a significant improvement over existing methods.

Recent progress has shown that large-scale pre-training using contrastive image-text pairs can be a promising alternative for high-quality visual representation learning from natural language supervision. Benefiting from a broader source of supervision, this new paradigm exhibits impressive transferability to downstream classification tasks and datasets. However, the problem of transferring the knowledge learned from image-text pairs to more complex dense prediction tasks has barely been visited. In this work, we present a new framework for dense prediction by implicitly and explicitly leveraging the pre-trained knowledge from CLIP. Specifically, we convert the original image-text matching problem in CLIP to a pixel-text matching problem and use the pixel-text score maps to guide the learning of dense prediction models. By further using the contextual information from the image to prompt the language model, we are able to facilitate our model to better exploit the pretrained knowledge. Our method is model-agnostic, which can be applied to arbitrary dense prediction systems and various pre-trained visual backbones including both CLIP models and ImageNet pre-trained models. Extensive experiments demonstrate the superior performance of our methods on semantic segmentation, object detection, and instance segmentation tasks. Code is available at https://github.com/raoyongming/DenseCLIP.

The resurgence of unsupervised learning can be attributed to the remarkable progress of self-supervised learning, which includes generative$(\mathcal{G})$and discriminative$(\mathcal{D})$models. In computer vision, the mainstream self-supervised learning algorithms are$\mathcal{D}$models. However, designing a$\mathcal{D}$model could be over-complicated; also, some studies hinted that a$\mathcal{D}$model might not be as general and interpretable as a$\mathcal{G}$model. In this paper, we switch from$\mathcal{D}$models to$\mathcal{G}$models using the classical auto-encoder$(AE)$. Note that a vanilla$\mathcal{G}$model was far less efficient than a$\mathcal{D}$model in self-supervised computer vision tasks, as it wastes model capability on overfitting semantic-agnostic high-frequency details. Inspired by perceptual learning that could use cross-view learning to perceive concepts and semantics11Following [26], we refer to semantics as visual concepts, e.g., a semantic-ware model indicates the model can perceive visual concepts, and the learned features are efficient in object recognition, detection, etc., we propose a novel$AE$that could learn semantic-aware representation via cross-view image reconstruction. We use one view of an image as the input and another view of the same image as the reconstruction target. This kind of$AE$has rarely been studied before, and the optimization is very difficult. To enhance learning ability and find a feasible solution, we propose a semantic aligner that uses geometric transformation knowledge to align the hidden code of$AE$to help optimization. These techniques significantly improve the representation learning ability of$AE$and make selfsupervised learning with$\mathcal{G}$models possible. Extensive experiments on many large-scale benchmarks (e.g., ImageNet, COCO 2017, and SYSU-30k) demonstrate the effectiveness of our methods. Code is available at https://github.com/wanggrun/Semantic-Aware-AE.

This paper presents a language-powered paradigm for ordinal regression. Existing methods usually treat each rank as a category and employ a set of weights to learn these concepts. These methods are easy to overfit and usually attain unsatisfactory performance as the learned concepts are mainly derived from the training set. Recent large pre-trained vision-language models like CLIP have shown impressive performance on various visual tasks. In this paper, we propose to learn the rank concepts from the rich semantic CLIP latent space. Specifically, we reformulate this task as an image-language matching problem with a contrastive objective, which regards labels as text and obtains a language prototype from a text encoder for each rank. While prompt engineering for CLIP is extremely time-consuming, we propose OrdinalCLIP, a differentiable prompting method for adapting CLIP for ordinal regression. OrdinalCLIP consists of learnable context tokens and learnable rank embeddings. The learnable rank embeddings are constructed by explicitly modeling numerical continuity, resulting in well-ordered, compact language prototypes in the CLIP space. Once learned, we can only save the language prototypes and discard the huge language model, resulting in zero additional computational overhead compared with the linear head counterpart. Experimental results show that our paradigm achieves competitive performance in general ordinal regression tasks, and gains improvements in few-shot and distribution shift settings for age estimation. The code is available at https://github.com/xk-huang/OrdinalCLIP.

Recent progress in vision Transformers exhibits great success in various tasks driven by the new spatial modeling mechanism based on dot-product self-attention. In this paper, we show that the key ingredients behind the vision Transformers, namely input-adaptive, long-range and high-order spatial interactions, can also be efficiently implemented with a convolution-based framework. We present the Recursive Gated Convolution ($\textit{g}^\textit{n}$Conv) that performs high-order spatial interactions with gated convolutions and recursive designs. The new operation is highly flexible and customizable, which is compatible with various variants of convolution and extends the two-order interactions in self-attention to arbitrary orders without introducing significant extra computation. $\textit{g}^\textit{n}$Conv can serve as a plug-and-play module to improve various vision Transformers and convolution-based models. Based on the operation, we construct a new family of generic vision backbones named HorNet. Extensive experiments on ImageNet classification, COCO object detection and ADE20K semantic segmentation show HorNet outperform Swin Transformers and ConvNeXt by a significant margin with similar overall architecture and training configurations. HorNet also shows favorable scalability to more training data and larger model sizes. Apart from the effectiveness in visual encoders, we also show $\textit{g}^\textit{n}$Conv can be applied to task-specific decoders and consistently improve dense prediction performance with less computation. Our results demonstrate that $\textit{g}^\textit{n}$Conv can be a new basic module for visual modeling that effectively combines the merits of both vision Transformers and CNNs. Code is available at https://github.com/raoyongming/HorNet.

In this paper, we investigate the problem of abductive visual reasoning (AVR), which requires vision systems to infer the most plausible explanation for visual observations. Unlike previous work which performs visual reasoning on static images or synthesized scenes, we exploit long-term reasoning from instructional videos that contain a wealth of detailed information about the physical world. We conceptualize two tasks for this emerging and challenging topic. The primary task is AVR, which is based on the initial configuration and desired goal from an instructional video, and the model is expected to figure out what is the most plausible sequence of steps to achieve the goal. In order to avoid trivial solutions based on appearance information rather than reasoning, the second task called AVR++ is constructed, which requires the model to answer why the unselected options are less plausible. We introduce a new dataset called VideoABC, which consists of 46,354 unique steps derived from 11,827 instructional videos, formulated as 13,526 abductive reasoning questions with an average reasoning duration of 51 seconds. Through an adversarial hard hypothesis mining algorithm, non-trivial and high-quality problems are generated efficiently and effectively. To achieve human-level reasoning, we propose a Hierarchical Dual Reasoning Network (HDRNet) to capture the long-term dependencies among steps and observations. We establish a benchmark for abductive visual reasoning, and our method set state-of-the-arts on AVR ( ∼ 74 %) and AVR++ ( ∼ 45 %), and humans can easily achieve over 90% accuracy on these two tasks. The large performance gap reveals the limitation of current video understanding models on temporal reasoning and leaves substantial room for future research on this challenging problem. Our dataset and code are available at https://github.com/wl-zhao/VideoABC.

Rapid progress and superior performance have been achieved for skeleton-based action recognition recently. In this article, we investigate this problem under a cross-dataset setting, which is a new, pragmatic, and challenging task in real-world scenarios. Following the unsupervised domain adaptation (UDA) paradigm, the action labels are only available on a source dataset, but unavailable on a target dataset in the training stage. Different from the conventional adversarial learning-based approaches for UDA, we utilize a self-supervision scheme to reduce the domain shift between two skeleton-based action datasets. Our inspiration is drawn from Cubism, an art genre from the early 20th century, which breaks and reassembles the objects to convey a greater context. By segmenting and permuting temporal segments or human body parts, we design two self-supervised learning classification tasks to explore the temporal and spatial dependency of a skeleton-based action and improve the generalization ability of the model. We conduct experiments on six datasets for skeleton-based action recognition, including three large-scale datasets (NTU RGB+D, PKU-MMD, and Kinetics) where new cross-dataset settings and benchmarks are established. Extensive results demonstrate that our method outperforms state-of-the-art approaches. The source codes of our model and all the compared methods are available at https://github.com/shanice-l/st-cubism.

Thanks to the substantial and explosively increased instructional videos on the Internet, novices are able to acquire knowledge for completing various tasks. Over the past decade, growing efforts have been devoted to investigating the problem on instructional video analysis. However, most existing instructional video datasets have limitations in diversity and scale, which makes them far from many real-world applications where more diverse activities occur. To address this, in this article, we propose a large-scale dataset called "COIN" for COmprehensive INstructional video analysis. Organized with a hierarchical structure, the COIN dataset contains 11,827 videos of 180 tasks in 12 domains (e.g., vehicles, gadgets, etc.) related to our daily life. With a new developed toolbox, all the videos are annotated efficiently with a series of step labels and the corresponding temporal boundaries. In order to provide a benchmark for instructional video analysis, we evaluate plenty of approaches on our COIN dataset under five different settings. Furthermore, we exploit two important characteristics (i.e., task-consistency and ordering-dependency) for localizing important steps in instructional videos. Accordingly, we propose two simple yet effective methods, which can be easily plugged into conventional proposal-based action detection models. We believe the introduction of the COIN dataset will promote the future in-depth research on instructional video analysis for the community. Our dataset, annotation toolbox and source codes are available at http://coin-dataset.github.io.

Assessing action quality from videos has attracted growing attention in recent years. Most existing approaches usually tackle this problem based on regression algorithms, which ignore the intrinsic ambiguity in the score labels caused by multiple judges or their subjective appraisals. To address this issue, we propose an uncertainty-aware score distribution learning (USDL) approach for action quality assessment (AQA). Specifically, we regard an action as an instance associated with a score distribution, which describes the probability of different evaluated scores. Moreover, under the circumstance where finer-grained score labels are available (e.g., difficulty degree of an action or multiple scores from different judges), we further devise a multi-path uncertainty-aware score distribution learning (MUSDL) method to explore the disentangled components of a score. In order to demonstrate the effectiveness of our proposed methods, We conduct experiments on two AQA datasets containing various Olympic actions. Our approaches set new state-of-the-arts under the Spearman's Rank Correlation (i.e., 0.8102 on AQA-7 and 0.9273 on MTL-AQA).

Recent years have witnessed rapid progress in employing graph convolutional networks (GCNs) for various video analysis tasks where graph-based data abound. However, exploring the transferable knowledge between different graphs, which is a direction with wide and potential applications, has been rarely studied. To address this issue, we propose a graph interaction networks (GINs) model for transferring relation knowledge across two graphs. Different from conventional domain adaptation or knowledge distillation approaches, our GINs focus on a “self-learned” weight matrix, which is a higher-level representation of the input data. And each element of the weight matrix represents the pair-wise relation among different nodes within the graph. Moreover, we guide the networks to transfer the knowledge across the weight matrices by designing a task-specific loss function, so that the relation information is well preserved during transfer. We conduct experiments on two different scenarios for video analysis, including a new proposed setting for unsupervised skeleton-based action recognition across different datasets, and supervised group activity recognition with multi-modal inputs. Extensive experiments on six widely used datasets illustrate that our GINs achieve very competitive performance in comparison with the state-of-the-arts.

There are substantial instruction videos on the Internet, which enables us to acquire knowledge for completing various tasks. However, most existing datasets for instruction video analysis have the limitations in diversity and scale, which makes them far from many real-world applications where more diverse activities occur. Moreover, it still remains a great challenge to organize and harness such data. To address these problems, we introduce a large-scale dataset called “COIN" for COmprehensive INstruction video analysis. Organized with a hierarchical structure, the COIN dataset contains 11,827 videos of 180 tasks in 12 domains (e.g., vehicles, gadgets, etc.) related to our daily life. With a new developed toolbox, all the videos are annotated effectively with a series of step descriptions and the corresponding temporal boundaries. Furthermore, we propose a simple yet effective method to capture the dependencies among different steps, which can be easily plugged into conventional proposal-based action detection methods for localizing important steps in instruction videos. In order to provide a benchmark for instruction video analysis, we evaluate plenty of approaches on the COIN dataset under different evaluation criteria. We expect the introduction of the COIN dataset will promote the future in-depth research on instruction video analysis for the community.

In this paper, we investigate the problem of RGB-D egocentric action recognition. Unlike conventional human action videos that are passively recorded by static cameras, egocentric videos are self-generated from wearable sensors that are more flexible and provide the close-ups with the visual attention of the wearers when they act. Moreover, RGB-D videos contain the spatial appearance and temporal information in the RGB modality and reflect the 3D structure of the scenes in the depth modality. To adequately learn the nonlinear structure of heterogeneous representations from different modalities and exploit their complementary characteristics, we develop a multi-stream deep neural networks (MDNN) method, which aims to preserve the distinctive property for each modality and simultaneously explore their sharable information in a unified deep architecture. Specifically, we deploy a Cauchy estimator to maximize the correlations of the sharable components and enforce the orthogonality constraints on the distinctive components to guarantee their high independencies. Since the egocentric action recognition is usually sensitive to hand poses, we extend our MDNN by integrating with the hand cues to enhance the recognition accuracy. Extensive experimental results on a newly collected data set and two additional benchmarks are presented to demonstrate the effectiveness of our proposed method for RGB-D egocentric action recognition.

In this paper, we investigate the problem of group activity recognition by learning semantics-perserving attention and contextual interaction among different people. Conventional methods usually aggregate the features extracted from individual persons by pooling operations, which lack physical meaning and cannot fully explore the contextual information for group activity recognition. To address this, we develop a Semantics-Preserving Teacher-Student (SPTS) networks architecture. Our SPTS networks first learn a Teacher Network in the semantic domain that classifies the word of group activity based on the words of individual actions. Then we design a Student Network in the appearance domain that recognizes the group activity according to the input video. We enforce the Student Network to mimic the Teacher Network in the learning procedure. In this way, we allocate semantics-preserving attention to different people, which is more effective to seek the key people and discard the misleading people, while no extra labelled data are required. Moreover, a group of people inherently lie in a graphbased structure, where the people and their relationship can be regarded as the nodes and edges of a graph respectively. Based on this, we build two graph convolutional modules on both the Teacher Network and the Student Network to reason the dependency among different people. Furthermore, we extend our approach on action segmentation task based on its intermediate features. Experimental results on four datasets for group activity analysis clearly show the superior performance of our method in comparisons with the state-of-the-arts.

In this paper, we propose a deep progressive reinforcement learning (DPRL) method for action recognition in skeleton-based videos, which aims to distil the most informative frames and discard ambiguous frames in sequences for recognizing actions. Since the choices of selecting representative frames are multitudinous for each video, we model the frame selection as a progressive process through deep reinforcement learning, during which we progressively adjust the chosen frames by taking two important factors into account: (1) the quality of the selected frames and (2) the relationship between the selected frames to the whole video. Moreover, considering the topology of human body inherently lies in a graph-based structure, where the vertices and edges represent the hinged joints and rigid bones respectively, we employ the graph-based convolutional neural network to capture the dependency between the joints for action recognition. Our approach achieves very competitive performance on three widely used benchmarks.

In this paper, we propose a Semantics-Preserving Teacher-Student (SPTS) model for group activity recognition in videos, which aims to mine the semantics-preserving attention to automatically seek the key people and discard the misleading people. Conventional methods usually aggregate the features extracted from individual persons by pooling operations, which cannot fully explore the contextual information for group activity recognition. To address this, our SPTS networks first learn a Teacher Network in semantic domain, which classifies the word of group activity based on the words of individual actions. Then we carefully design a Student Network in vision domain, which recognizes the group activity according to the input videos, and enforce the Student Network to mimic the Teacher Network during the learning process. In this way, we allocate semantics-preserving attention to different people, which adequately explores the contextual information of different people and requires no extra labelled data. Experimental results on two widely used benchmarks for group activity recognition clearly show the superior performance of our method in comparisons with the state-of-the-arts.

In this paper, we investigate the problem of action recognition in RGB-D egocentric videos. These self-generated and embodied videos provide richer semantic cues than the conventional videos captured from the third-person view for action recognition. Moreover, they contain both appearance information and 3D structure of the scenes from the RGB modality and depth modality respectively. Motivated by these advantages, we first collect a video-based RGB-D egocentric dataset (THU-READ) with diverse types of daily-life actions. Then we evaluate several approaches including hand-crafted features and deep learning methods on THU-READ. To improve the performance, we further develop a tri-stream convolutional network (TCNet) method, which learns to exploit the fuse with both the RGB and depth modalities for action recognition. Experimental results show that our model achieves competitive performance with state-of-the-art methods.