The third Pixel-level Video Understanding in the Wild (PVUW CVPR 2024) challenge aims to advance the state of art in video understanding through benchmarking Video Panoptic Segmentation (VPS) and Video Semantic Segmentation (VSS) on challenging videos and scenes introduced in the large-scale Video Panoptic Segmentation in the Wild (VIPSeg) test set and the large-scale Video Scene Parsing in the Wild (VSPW) test set, respectively. This paper details our research work that achieved the 1st place winner in the PVUW'24 VPS challenge, establishing state of art results in all metrics, including the Video Panoptic Quality (VPQ) and Segmentation and Tracking Quality (STQ). With minor fine-tuning our approach also achieved the 3rd place in the PVUW'24 VSS challenge ranked by the mIoU (mean intersection over union) metric and the first place ranked by the VC16 (16-frame video consistency) metric. Our winning solution stands on the shoulders of giant foundational vision transformer model (DINOv2 ViT-g) and proven multi-stage Decoupled Video Instance Segmentation (DVIS) frameworks for video understanding.
Pixel-level Video Understanding requires effectively integrating three-dimensional data in both spatial and temporal dimensions to learn accurate and stable semantic information from continuous frames. However, existing advanced models on the VSPW dataset have not fully modeled spatiotemporal relationships. In this paper, we present our solution for the PVUW competition, where we introduce masked video consistency (MVC) based on existing models. MVC enforces the consistency between predictions of masked frames where random patches are withheld. The model needs to learn the segmentation results of the masked parts through the context of images and the relationship between preceding and succeeding frames of the video. Additionally, we employed test-time augmentation, model aggeregation and a multimodal model-based post-processing method. Our approach achieves 67.27% mIoU performance on the VSPW dataset, ranking 2nd place in the PVUW2024 challenge VSS track.
We present the ShareGPT4Video series, aiming to facilitate the video understanding of large video-language models (LVLMs) and the video generation of text-to-video models (T2VMs) via dense and precise captions. The series comprises: 1) ShareGPT4Video, 40K GPT4V annotated dense captions of videos with various lengths and sources, developed through carefully designed data filtering and annotating strategy. 2) ShareCaptioner-Video, an efficient and capable captioning model for arbitrary videos, with 4.8M high-quality aesthetic videos annotated by it. 3) ShareGPT4Video-8B, a simple yet superb LVLM that reached SOTA performance on three advancing video benchmarks. To achieve this, taking aside the non-scalable costly human annotators, we find using GPT4V to caption video with a naive multi-frame or frame-concatenation input strategy leads to less detailed and sometimes temporal-confused results. We argue the challenge of designing a high-quality video captioning strategy lies in three aspects: 1) Inter-frame precise temporal change understanding. 2) Intra-frame detailed content description. 3) Frame-number scalability for arbitrary-length videos. To this end, we meticulously designed a differential video captioning strategy, which is stable, scalable, and efficient for generating captions for videos with arbitrary resolution, aspect ratios, and length. Based on it, we construct ShareGPT4Video, which contains 40K high-quality videos spanning a wide range of categories, and the resulting captions encompass rich world knowledge, object attributes, camera movements, and crucially, detailed and precise temporal descriptions of events. Based on ShareGPT4Video, we further develop ShareCaptioner-Video, a superior captioner capable of efficiently generating high-quality captions for arbitrary videos...
Video panoptic segmentation is an advanced task that extends panoptic segmentation by applying its concept to video sequences. In the hope of addressing the challenge of video panoptic segmentation in diverse conditions, We utilize DVIS++ as our baseline model and enhance it by introducing a comprehensive approach centered on the query-wise ensemble, supplemented by additional techniques. Our proposed approach achieved a VPQ score of 57.01 on the VIPSeg test set, and ranked 3rd in the VPS track of the 3rd Pixel-level Video Understanding in the Wild Challenge.
The evaluation of Long Video Understanding (LVU) performance poses an important but challenging research problem. Despite previous efforts, the existing video understanding benchmarks are severely constrained by several issues, especially the insufficient lengths of videos, a lack of diversity in video types and evaluation tasks, and the inappropriateness for evaluating LVU performances. To address the above problems, we propose a new benchmark called MLVU (Multi-task Long Video Understanding Benchmark) for the comprehensive and in-depth evaluation of LVU. MLVU presents the following critical values: \textit{1)} The substantial and flexible extension of video lengths, which enables the benchmark to evaluate LVU performance across a wide range of durations. \textit{2)} The inclusion of various video genres, e.g., movies, surveillance footage, egocentric videos, cartoons, game videos, etc., which reflects the models' LVU performances in different scenarios. \textit{3)} The development of diversified evaluation tasks, which enables a comprehensive examination of MLLMs' key abilities in long-video understanding. The empirical study with 23 latest MLLMs reveals significant room for improvement in today's technique, as all existing methods struggle with most of the evaluation tasks and exhibit severe performance degradation when handling longer videos. Additionally, it suggests that factors such as context length, image-understanding ability, and the choice of LLM backbone can play critical roles in future advancements. We anticipate that MLVU will advance the research of long video understanding by providing a comprehensive and in-depth analysis of MLLMs.
We introduce LAVITI, a novel approach to learning language, video, and temporal representations in long-form videos via contrastive learning. Different from pre-training on video-text pairs like EgoVLP, LAVITI aims to align language, video, and temporal features by extracting meaningful moments in untrimmed videos. Our model employs a set of learnable moment queries to decode clip-level visual, language, and temporal features. In addition to vision and language alignment, we introduce relative temporal embeddings (TE) to represent timestamps in videos, which enables contrastive learning of time. Significantly different from traditional approaches, the prediction of a particular timestamp is transformed by computing the similarity score between the predicted TE and all TEs. Furthermore, existing approaches for video understanding are mainly designed for short videos due to high computational complexity and memory footprint. Our method can be trained on the Ego4D dataset with only 8 NVIDIA RTX-3090 GPUs in a day. We validated our method on CharadesEgo action recognition, achieving state-of-the-art results.
Video Panoptic Segmentation (VPS) is a challenging task that is extends from image panoptic segmentation.VPS aims to simultaneously classify, track, segment all objects in a video, including both things and stuff. Due to its wide application in many downstream tasks such as video understanding, video editing, and autonomous driving. In order to deal with the task of video panoptic segmentation in the wild, we propose a robust integrated video panoptic segmentation solution. We use DVIS++ framework as our baseline to generate the initial masks. Then,we add an additional image semantic segmentation model to further improve the performance of semantic classes.Finally, our method achieves state-of-the-art performance with a VPQ score of 56.36 and 57.12 in the development and test phases, respectively, and ultimately ranked 2nd in the VPS track of the PVUW Challenge at CVPR2024.
Current video-language models (VLMs) rely extensively on instance-level alignment between video and language modalities, which presents two major limitations: (1) visual reasoning disobeys the natural perception that humans do in first-person perspective, leading to a lack of reasoning interpretation; and (2) learning is limited in capturing inherent fine-grained relationships between two modalities. In this paper, we take an inspiration from human perception and explore a compositional approach for egocentric video representation. We introduce HENASY (Hierarchical ENtities ASsemblY), which includes a spatiotemporal token grouping mechanism to explicitly assemble dynamically evolving scene entities through time and model their relationship for video representation. By leveraging compositional structure understanding, HENASY possesses strong interpretability via visual grounding with free-form text queries. We further explore a suite of multi-grained contrastive losses to facilitate entity-centric understandings. This comprises three alignment types: video-narration, noun-entity, verb-entities alignments. Our method demonstrates strong interpretability in both quantitative and qualitative experiments; while maintaining competitive performances on five downstream tasks via zero-shot transfer or as video/text representation, including video/text retrieval, action recognition, multi-choice query, natural language query, and moments query.
Procedural activities are sequences of key-steps aimed at achieving specific goals. They are crucial to build intelligent agents able to assist users effectively. In this context, task graphs have emerged as a human-understandable representation of procedural activities, encoding a partial ordering over the key-steps. While previous works generally relied on hand-crafted procedures to extract task graphs from videos, in this paper, we propose an approach based on direct maximum likelihood optimization of edges' weights, which allows gradient-based learning of task graphs and can be naturally plugged into neural network architectures. Experiments on the CaptainCook4D dataset demonstrate the ability of our approach to predict accurate task graphs from the observation of action sequences, with an improvement of +16.7% over previous approaches. Owing to the differentiability of the proposed framework, we also introduce a feature-based approach, aiming to predict task graphs from key-step textual or video embeddings, for which we observe emerging video understanding abilities. Task graphs learned with our approach are also shown to significantly enhance online mistake detection in procedural egocentric videos, achieving notable gains of +19.8% and +7.5% on the Assembly101-O and EPIC-Tent-O datasets. Code for replicating experiments is available at https://github.com/fpv-iplab/Differentiable-Task-Graph-Learning.
Recently, video generation techniques have advanced rapidly. Given the popularity of video content on social media platforms, these models intensify concerns about the spread of fake information. Therefore, there is a growing demand for detectors capable of distinguishing between fake AI-generated videos and mitigating the potential harm caused by fake information. However, the lack of large-scale datasets from the most advanced video generators poses a barrier to the development of such detectors. To address this gap, we introduce the first AI-generated video detection dataset, GenVideo. It features the following characteristics: (1) a large volume of videos, including over one million AI-generated and real videos collected; (2) a rich diversity of generated content and methodologies, covering a broad spectrum of video categories and generation techniques. We conducted extensive studies of the dataset and proposed two evaluation methods tailored for real-world-like scenarios to assess the detectors' performance: the cross-generator video classification task assesses the generalizability of trained detectors on generators; the degraded video classification task evaluates the robustness of detectors to handle videos that have degraded in quality during dissemination. Moreover, we introduced a plug-and-play module, named Detail Mamba (DeMamba), designed to enhance the detectors by identifying AI-generated videos through the analysis of inconsistencies in temporal and spatial dimensions. Our extensive experiments demonstrate DeMamba's superior generalizability and robustness on GenVideo compared to existing detectors. We believe that the GenVideo dataset and the DeMamba module will significantly advance the field of AI-generated video detection. Our code and dataset will be aviliable at \url{https://github.com/chenhaoxing/DeMamba}.
Surgical phase recognition has gained significant attention due to its potential to offer solutions to numerous demands of the modern operating room. However, most existing methods concentrate on minimally invasive surgery (MIS), leaving surgical phase recognition for open surgery understudied. This discrepancy is primarily attributed to the scarcity of publicly available open surgery video datasets for surgical phase recognition. To address this issue, we introduce a new egocentric open surgery video dataset for phase recognition, named EgoSurgery-Phase. This dataset comprises 15 hours of real open surgery videos spanning 9 distinct surgical phases all captured using an egocentric camera attached to the surgeon's head. In addition to video, the EgoSurgery-Phase offers eye gaze. As far as we know, it is the first real open surgery video dataset for surgical phase recognition publicly available. Furthermore, inspired by the notable success of masked autoencoders (MAEs) in video understanding tasks (e.g., action recognition), we propose a gaze-guided masked autoencoder (GGMAE). Considering the regions where surgeons' gaze focuses are often critical for surgical phase recognition (e.g., surgical field), in our GGMAE, the gaze information acts as an empirical semantic richness prior to guiding the masking process, promoting better attention to semantically rich spatial regions. GGMAE significantly improves the previous state-of-the-art recognition method (6.4% in Jaccard) and the masked autoencoder-based method (3.1% in Jaccard) on EgoSurgery-Phase. The dataset is released at https://github.com/Fujiry0/EgoSurgery.
Long-form video understanding has been a challenging task due to the high redundancy in video data and the abundance of query-irrelevant information. To tackle this challenge, we propose VideoTree, a training-free framework which builds a query-adaptive and hierarchical video representation for LLM reasoning over long-form videos. First, VideoTree extracts query-relevant information from the input video through an iterative process, progressively refining the selection of keyframes based on their relevance to the query. Furthermore, VideoTree leverages the inherent hierarchical structure of long video data, which is often overlooked by existing LLM-based methods. Specifically, we incorporate multigranularity information into a tree-based representation, allowing VideoTree to extract query-relevant details from long videos in a coarse-to-fine manner. This enables the model to effectively handle a wide range of video queries with varying levels of detail. Finally, VideoTree aggregates the hierarchical query-relevant information within the tree structure and feeds it into an LLM reasoning model to answer the query. Our experiments show that our training-free method improves both reasoning accuracy and efficiency compared to existing methods. Specifically, VideoTree outperforms the existing training-free approaches on the popular EgoSchema and NExT-QA benchmarks with less inference time, achieving 61.1% and 75.6% accuracy on the test set without additional video-specific training. Moreover, on the long split of Video-MME benchmark (average 44 minutes), the training-free VideoTree framework achieves better performance than the strong proprietary GPT-4V model and other MLLMs that were extensively trained on video data.
Video recognition remains an open challenge, requiring the identification of diverse content categories within videos. Mainstream approaches often perform flat classification, overlooking the intrinsic hierarchical structure relating categories. To address this, we formalize the novel task of hierarchical video recognition, and propose a video-language learning framework tailored for hierarchical recognition. Specifically, our framework encodes dependencies between hierarchical category levels, and applies a top-down constraint to filter recognition predictions. We further construct a new fine-grained dataset based on medical assessments for rehabilitation of stroke patients, serving as a challenging benchmark for hierarchical recognition. Through extensive experiments, we demonstrate the efficacy of our approach for hierarchical recognition, significantly outperforming conventional methods, especially for fine-grained subcategories. The proposed framework paves the way for hierarchical modeling in video understanding tasks, moving beyond flat categorization.
Recent advancements in Multi-modal Large Language Models (MLLMs) have significantly improved their performance in tasks combining vision and language. However, challenges persist in detailed multi-modal understanding, comprehension of complex tasks, and reasoning over multi-modal information. This paper introduces MMCTAgent, a novel multi-modal critical thinking agent framework designed to address the inherent limitations of current MLLMs in complex visual reasoning tasks. Inspired by human cognitive processes and critical thinking, MMCTAgent iteratively analyzes multi-modal information, decomposes queries, plans strategies, and dynamically evolves its reasoning. Additionally, MMCTAgent incorporates critical thinking elements such as verification of final answers and self-reflection through a novel approach that defines a vision-based critic and identifies task-specific evaluation criteria, thereby enhancing its decision-making abilities. Through rigorous evaluations across various image and video understanding benchmarks, we demonstrate that MMCTAgent (with and without the critic) outperforms both foundational MLLMs and other tool-augmented pipelines.
This paper presents VideoStreaming, an advanced vision-language large model (VLLM) for video understanding, that capably understands arbitrary-length video with a constant number of video tokens streamingly encoded and adaptively selected. The challenge of video understanding in the vision language area mainly lies in the significant computational burden caused by the great number of tokens extracted from long videos. Previous works rely on sparse sampling or frame compression to reduce tokens. However, such approaches either disregard temporal information in a long time span or sacrifice spatial details, resulting in flawed compression. To address these limitations, our VideoStreaming has two core designs: Memory-Propagated Streaming Encoding and Adaptive Memory Selection. The Memory-Propagated Streaming Encoding architecture segments long videos into short clips and sequentially encodes each clip with a propagated memory. In each iteration, we utilize the encoded results of the preceding clip as historical memory, which is integrated with the current clip to distill a condensed representation that encapsulates the video content up to the current timestamp. After the encoding process, the Adaptive Memory Selection strategy selects a constant number of question-related memories from all the historical memories and feeds them into the LLM to generate informative responses. The question-related selection reduces redundancy within the memories, enabling efficient and precise video understanding. Meanwhile, the disentangled video extraction and reasoning design allows the LLM to answer different questions about a video by directly selecting corresponding memories, without the need to encode the whole video for each question. Our model achieves superior performance and higher efficiency on long video benchmarks, showcasing precise temporal comprehension for detailed question answering.
Video Anomaly Detection (VAD) systems can autonomously monitor and identify disturbances, reducing the need for manual labor and associated costs. However, current VAD systems are often limited by their superficial semantic understanding of scenes and minimal user interaction. Additionally, the prevalent data scarcity in existing datasets restricts their applicability in open-world scenarios. In this paper, we introduce Hawk, a novel framework that leverages interactive large Visual Language Models (VLM) to interpret video anomalies precisely. Recognizing the difference in motion information between abnormal and normal videos, Hawk explicitly integrates motion modality to enhance anomaly identification. To reinforce motion attention, we construct an auxiliary consistency loss within the motion and video space, guiding the video branch to focus on the motion modality. Moreover, to improve the interpretation of motion-to-language, we establish a clear supervisory relationship between motion and its linguistic representation. Furthermore, we have annotated over 8,000 anomaly videos with language descriptions, enabling effective training across diverse open-world scenarios, and also created 8,000 question-answering pairs for users' open-world questions. The final results demonstrate that Hawk achieves SOTA performance, surpassing existing baselines in both video description generation and question-answering. Our codes/dataset/demo will be released at https://github.com/jqtangust/hawk.
Point cloud videos can faithfully capture real-world spatial geometries and temporal dynamics, which are essential for enabling intelligent agents to understand the dynamically changing world. However, designing an effective 4D backbone remains challenging, mainly due to the irregular and unordered distribution of points and temporal inconsistencies across frames. Also, recent transformer-based 4D backbones commonly suffer from large computational costs due to their quadratic complexity, particularly for long video sequences.To address these challenges, we propose a novel point cloud video understanding backbone purely based on the State Space Models (SSMs). Specifically, we first disentangle space and time in 4D video sequences and then establish the spatio-temporal correlation with our designed Mamba blocks. The Intra-frame Spatial Mamba module is developed to encode locally similar geometric structures within a certain temporal stride. Subsequently, locally correlated tokens are delivered to the Inter-frame Temporal Mamba module, which integrates long-term point features across the entire video with linear complexity. Our proposed Mamba4d achieves competitive performance on the MSR-Action3D action recognition (+10.4% accuracy), HOI4D action segmentation (+0.7 F1 Score), and Synthia4D semantic segmentation (+0.19 mIoU) datasets. Especially, for long video sequences, our method has a significant efficiency improvement with 87.5% GPU memory reduction and 5.36 times speed-up.
Recent advancements in image understanding have benefited from the extensive use of web image-text pairs. However, video understanding remains a challenge despite the availability of substantial web video-text data. This difficulty primarily arises from the inherent complexity of videos and the inefficient language supervision in recent web-collected video-text datasets. In this paper, we introduce Text-Only Pre-Alignment (TOPA), a novel approach to extend large language models (LLMs) for video understanding, without the need for pre-training on real video data. Specifically, we first employ an advanced LLM to automatically generate Textual Videos comprising continuous textual frames, along with corresponding annotations to simulate real video-text data. Then, these annotated textual videos are used to pre-align a language-only LLM with the video modality. To bridge the gap between textual and real videos, we employ the CLIP model as the feature extractor to align image and text modalities. During text-only pre-alignment, the continuous textual frames, encoded as a sequence of CLIP text features, are analogous to continuous CLIP image features, thus aligning the LLM with real video representation. Extensive experiments, including zero-shot evaluation and finetuning on various video understanding tasks, demonstrate that TOPA is an effective and efficient framework for aligning video content with LLMs. In particular, without training on any video data, the TOPA-Llama2-13B model achieves a Top-1 accuracy of 51.0% on the challenging long-form video understanding benchmark, Egoschema. This performance surpasses previous video-text pre-training approaches and proves competitive with recent GPT-3.5-based video agents.
Do we fully leverage the potential of visual encoder in Multimodal Large Language Models (MLLMs)? The recent outstanding performance of MLLMs in multimodal understanding has garnered broad attention from both academia and industry. In the current MLLM rat race, the focus seems to be predominantly on the linguistic side. We witness the rise of larger and higher-quality instruction datasets, as well as the involvement of larger-sized LLMs. Yet, scant attention has been directed towards the visual signals utilized by MLLMs, often assumed to be the final high-level features extracted by a frozen visual encoder. In this paper, we introduce the Dense Connector - a simple, effective, and plug-and-play vision-language connector that significantly enhances existing MLLMs by leveraging multi-layer visual features, with minimal additional computational overhead. Building on this, we also propose the Efficient Dense Connector, which achieves performance comparable to LLaVA-v1.5 with only 25% of the visual tokens. Furthermore, our model, trained solely on images, showcases remarkable zero-shot capabilities in video understanding as well. Experimental results across various vision encoders, image resolutions, training dataset scales, varying sizes of LLMs (2.7B->70B), and diverse architectures of MLLMs (e.g., LLaVA-v1.5, LLaVA-NeXT and Mini-Gemini) validate the versatility and scalability of our approach, achieving state-of-the-art performance across 19 image and video benchmarks. We hope that this work will provide valuable experience and serve as a basic module for future MLLM development. Code is available at https://github.com/HJYao00/DenseConnector .
In this work, we introduce (a) the new problem of anticipating object state changes in images and videos during procedural activities, (b) new curated annotation data for object state change classification based on the Ego4D dataset, and (c) the first method for addressing this challenging problem. Solutions to this new task have important implications in vision-based scene understanding, automated monitoring systems, and action planning. The proposed novel framework predicts object state changes that will occur in the near future due to yet unseen human actions by integrating learned visual features that represent recent visual information with natural language (NLP) features that represent past object state changes and actions. Leveraging the extensive and challenging Ego4D dataset which provides a large-scale collection of first-person perspective videos across numerous interaction scenarios, we introduce an extension noted Ego4D-OSCA that provides new curated annotation data for the object state change anticipation task (OSCA). An extensive experimental evaluation is presented demonstrating the proposed method's efficacy in predicting object state changes in dynamic scenarios. The performance of the proposed approach also underscores the potential of integrating video and linguistic cues to enhance the predictive performance of video understanding systems and lays the groundwork for future research on the new task of object state change anticipation. The source code and the new annotation data (Ego4D-OSCA) will be made publicly available.