With the growth of high-quality data and advancement in visual pre-training paradigms, Video Foundation Models (VFMs) have made significant progress recently, demonstrating their remarkable performance on traditional video understanding benchmarks. However, the existing benchmarks (e.g. Kinetics) and their evaluation protocols are often limited by relatively poor diversity, high evaluation costs, and saturated performance metrics. In this paper, we build a comprehensive benchmark suite to address these issues, namely VideoEval. Specifically, we establish the Video Task Adaption Benchmark (VidTAB) and the Video Embedding Benchmark (VidEB) from two perspectives: evaluating the task adaptability of VFMs under few-shot conditions and assessing their representation power by directly applying to downstream tasks. With VideoEval, we conduct a large-scale study on 20 popular open-source vision foundation models. Our study reveals some insightful findings on VFMs: 1) overall, current VFMs exhibit weak generalization across diverse tasks, 2) increasing video data, whether labeled or weakly-labeled video-text pairs, does not necessarily improve task performance, 3) the effectiveness of some pre-training paradigms may not be fully validated in previous benchmarks, and 4) combining different pre-training paradigms can help improve the generalization capabilities. We believe this study serves as an important complement to the current evaluation for VFMs and offers valuable insights for the future research.
Image-to-video adaptation seeks to efficiently adapt image models for use in the video domain. Instead of finetuning the entire image backbone, many image-to-video adaptation paradigms use lightweight adapters for temporal modeling on top of the spatial module. However, these attempts are subject to limitations in efficiency and interpretability. In this paper, we propose a novel and efficient image-to-video adaptation strategy from the object-centric perspective. Inspired by human perception, which identifies objects as key components for video understanding, we integrate a proxy task of object discovery into image-to-video transfer learning. Specifically, we adopt slot attention with learnable queries to distill each frame into a compact set of object tokens. These object-centric tokens are then processed through object-time interaction layers to model object state changes across time. Integrated with two novel object-level losses, we demonstrate the feasibility of performing efficient temporal reasoning solely on the compressed object-centric representations for video downstream tasks. Our method achieves state-of-the-art performance with fewer tunable parameters, only 5\% of fully finetuned models and 50\% of efficient tuning methods, on action recognition benchmarks. In addition, our model performs favorably in zero-shot video object segmentation without further retraining or object annotations, proving the effectiveness of object-centric video understanding.
Human body actions are an important form of non-verbal communication in social interactions. This paper specifically focuses on a subset of body actions known as micro-actions, which are subtle, low-intensity body movements with promising applications in human emotion analysis. In real-world scenarios, human micro-actions often temporally co-occur, with multiple micro-actions overlapping in time, such as concurrent head and hand movements. However, current research primarily focuses on recognizing individual micro-actions while overlooking their co-occurring nature. To address this gap, we propose a new task named Multi-label Micro-Action Detection (MMAD), which involves identifying all micro-actions in a given short video, determining their start and end times, and categorizing them. Accomplishing this requires a model capable of accurately capturing both long-term and short-term action relationships to detect multiple overlapping micro-actions. To facilitate the MMAD task, we introduce a new dataset named Multi-label Micro-Action-52 (MMA-52) and propose a baseline method equipped with a dual-path spatial-temporal adapter to address the challenges of subtle visual change in MMAD. We hope that MMA-52 can stimulate research on micro-action analysis in videos and prompt the development of spatio-temporal modeling in human-centric video understanding. The proposed MMA-52 dataset is available at: https://github.com/VUT-HFUT/Micro-Action.
We introduce OmChat, a model designed to excel in handling long contexts and video understanding tasks. OmChat's new architecture standardizes how different visual inputs are processed, making it more efficient and adaptable. It uses a dynamic vision encoding process to effectively handle images of various resolutions, capturing fine details across a range of image qualities. OmChat utilizes an active progressive multimodal pretraining strategy, which gradually increases the model's capacity for long contexts and enhances its overall abilities. By selecting high-quality data during training, OmChat learns from the most relevant and informative data points. With support for a context length of up to 512K, OmChat demonstrates promising performance in tasks involving multiple images and videos, outperforming most open-source models in these benchmarks. Additionally, OmChat proposes a prompting strategy for unifying complex multimodal inputs including single image text, multi-image text and videos, and achieving competitive performance on single-image benchmarks. To further evaluate the model's capabilities, we proposed a benchmark dataset named Temporal Visual Needle in a Haystack. This dataset assesses OmChat's ability to comprehend temporal visual details within long videos. Our analysis highlights several key factors contributing to OmChat's success: support for any-aspect high image resolution, the active progressive pretraining strategy, and high-quality supervised fine-tuning datasets. This report provides a detailed overview of OmChat's capabilities and the strategies that enhance its performance in visual understanding.
The performance of Large Vision Language Models (LVLMs) is dependent on the size and quality of their training datasets. Existing video instruction tuning datasets lack diversity as they are derived by prompting large language models with video captions to generate question-answer pairs, and are therefore mostly descriptive. Meanwhile, many labeled video datasets with diverse labels and supervision exist - however, we find that their integration into LVLMs is non-trivial. Herein, we present Video Self-Training with augmented Reasoning (Video-STaR), the first video self-training approach. Video-STaR allows the utilization of any labeled video dataset for video instruction tuning. In Video-STaR, an LVLM cycles between instruction generation and finetuning, which we show (I) improves general video understanding and (II) adapts LVLMs to novel downstream tasks with existing supervision. During generation, an LVLM is prompted to propose an answer. The answers are then filtered only to those that contain the original video labels, and the LVLM is then re-trained on the generated dataset. By only training on generated answers that contain the correct video labels, Video-STaR utilizes these existing video labels as weak supervision for video instruction tuning. Our results demonstrate that Video-STaR-enhanced LVLMs exhibit improved performance in (I) general video QA, where TempCompass performance improved by 10%, and (II) on downstream tasks, where Video-STaR improved Kinetics700-QA accuracy by 20% and action quality assessment on FineDiving by 15%.
We present InternLM-XComposer-2.5 (IXC-2.5), a versatile large-vision language model that supports long-contextual input and output. IXC-2.5 excels in various text-image comprehension and composition applications, achieving GPT-4V level capabilities with merely 7B LLM backend. Trained with 24K interleaved image-text contexts, it can seamlessly extend to 96K long contexts via RoPE extrapolation. This long-context capability allows IXC-2.5 to excel in tasks requiring extensive input and output contexts. Compared to its previous 2.0 version, InternLM-XComposer-2.5 features three major upgrades in vision-language comprehension: (1) Ultra-High Resolution Understanding, (2) Fine-Grained Video Understanding, and (3) Multi-Turn Multi-Image Dialogue. In addition to comprehension, IXC-2.5 extends to two compelling applications using extra LoRA parameters for text-image composition: (1) Crafting Webpages and (2) Composing High-Quality Text-Image Articles. IXC-2.5 has been evaluated on 28 benchmarks, outperforming existing open-source state-of-the-art models on 16 benchmarks. It also surpasses or competes closely with GPT-4V and Gemini Pro on 16 key tasks. The InternLM-XComposer-2.5 is publicly available at https://github.com/InternLM/InternLM-XComposer.
Recently, with the rise of web videos, managing and understanding large-scale video datasets has become increasingly important. Video Large Language Models (VideoLLMs) have emerged in recent years due to their strong video understanding capabilities. However, training and inference processes for VideoLLMs demand vast amounts of data, presenting significant challenges to data management, particularly regarding efficiency, robustness, and effectiveness. In this work, we present KeyVideoLLM, a text-video frame similarity-based keyframe selection method designed to manage VideoLLM data efficiently, robustly, and effectively. Specifically, KeyVideoLLM achieves a remarkable data compression rate of up to 60.9 times, substantially lowering disk space requirements, which proves its high efficiency. Additionally, it maintains a 100% selection success rate across all video formats and scales, enhances processing speed by up to 200 times compared to existing keyframe selection methods, and does not require hyperparameter tuning. Beyond its outstanding efficiency and robustness, KeyVideoLLM further improves model performance in video question-answering tasks during both training and inference stages. Notably, it consistently achieved the state-of-the-art (SoTA) experimental results on diverse datasets.
The advent of video-based Large Language Models (LLMs) has significantly enhanced video understanding. However, it has also raised some safety concerns regarding data protection, as videos can be more easily annotated, even without authorization. This paper introduces Video Watermarking, a novel technique to protect videos from unauthorized annotations by such video-based LLMs, especially concerning the video content and description, in response to specific queries. By imperceptibly embedding watermarks into key video frames with multi-modal flow-based losses, our method preserves the viewing experience while preventing misuse by video-based LLMs. Extensive experiments show that Video Watermarking significantly reduces the comprehensibility of videos with various video-based LLMs, demonstrating both stealth and robustness. In essence, our method provides a solution for securing video content, ensuring its integrity and confidentiality in the face of evolving video-based LLMs technologies.
Generating fine-grained video descriptions is a fundamental challenge in video understanding. In this work, we introduce Tarsier, a family of large-scale video-language models designed to generate high-quality video descriptions. Tarsier employs CLIP-ViT to encode frames separately and then uses an LLM to model temporal relationships. Despite its simple architecture, we demonstrate that with a meticulously designed two-stage training procedure, the Tarsier models exhibit substantially stronger video description capabilities than any existing open-source model, showing a +51.4% advantage in human side-by-side evaluation over the strongest model. Additionally, they are comparable to state-of-the-art proprietary models, with a +12.3% advantage against GPT-4V and a −6.7% disadvantage against Gemini 1.5 Pro. Besides video description, Tarsier proves to be a versatile generalist model, achieving new state-of-the-art results across nine public benchmarks, including multi-choice VQA, open-ended VQA, and zero-shot video captioning. Our second contribution is the introduction of a new benchmark for evaluating video description models, consisting of a new challenging dataset featuring videos from diverse sources and varying complexity, along with an automatic method specifically designed to assess the quality of fine-grained video descriptions. We make our models and evaluation benchmark publicly available at https://github.com/bytedance/tarsier (coming soon).
Understanding long videos, ranging from tens of minutes to several hours, presents unique challenges in video comprehension. Despite the increasing importance of long-form video content, existing benchmarks primarily focus on shorter clips. To address this gap, we introduce InfiniBench a comprehensive benchmark for very long video understanding which presents 1)The longest video duration, averaging 52.59 minutes per video 2) The largest number of question-answer pairs, 108.2K 3) Diversity in questions that examine nine different skills and include both multiple-choice questions and open-ended questions 4) Human-centric, as the video sources come from movies and daily TV shows, with specific human-level question designs such as Movie Spoiler Questions that require critical thinking and comprehensive understanding. Using InfiniBench, we comprehensively evaluate existing Large Multi-Modality Models (LMMs) on each skill, including the commercial models such as GPT-4o and Gemini 1.5 Flash and the open-source models. The evaluation shows significant challenges in our benchmark. Our findings reveal that even leading AI models like GPT-4o and Gemini 1.5 Flash face challenges in achieving high performance in long video understanding, with average accuracies of just 49.16\% and 42.72\%, and average scores of 3.22 and 2.71 out of 5, respectively. We hope this benchmark will stimulate the LMMs community towards long video and human-level understanding. Our benchmark can be accessed at https://vision-cair.github.io/InfiniBench/
Video understanding requires the extraction of rich spatio-temporal representations, which transformer models achieve through self-attention. Unfortunately, self-attention poses a computational burden. In NLP, Mamba has surfaced as an efficient alternative for transformers. However, Mamba's successes do not trivially extend to vision tasks, including those in video analysis. In this paper, we theoretically analyze the differences between self-attention and Mamba. We identify two limitations in Mamba's token processing: historical decay and element contradiction. We propose VideoMambaPro (VMP) that solves the identified limitations by adding masked backward computation and elemental residual connections to a VideoMamba backbone. Differently sized VideoMambaPro models surpass VideoMamba by 1.6-2.8% and 1.1-1.9% top-1 on Kinetics-400 and Something-Something V2, respectively. Even without extensive pre-training, our models present an increasingly attractive and efficient alternative to current transformer models. Moreover, our two solutions are orthogonal to recent advances in Vision Mamba models, and are likely to provide further improvements in future models.
Transformer architectures such as Vision Transformers (ViT) have proven effective for solving visual perception tasks. However, they suffer from two major limitations; first, the quadratic complexity of self-attention limits the number of tokens that can be processed, and second, Transformers often require large amounts of training data to attain state-of-the-art performance. In this paper, we propose a new multi-head self-attention (MHSA) variant named Fibottention, which can replace MHSA in Transformer architectures. Fibottention is data-efficient and computationally more suitable for processing large numbers of tokens than the standard MHSA. It employs structured sparse attention based on dilated Fibonacci sequences, which, uniquely, differ across attention heads, resulting in inception-like diverse features across heads. The spacing of the Fibonacci sequences follows the Wythoff array, which minimizes the redundancy of token interactions aggregated across different attention heads, while still capturing sufficient complementary information through token pair interactions. These sparse attention patterns are unique among the existing sparse attention and lead to an $O(N \log N)$ complexity, where $N$ is the number of tokens. Leveraging only 2-6% of the elements in the self-attention heads, Fibottention embedded into popular, state-of-the-art Transformer architectures can achieve significantly improved predictive performance for domains with limited data such as image classification, video understanding, and robot learning tasks, and render reduced computational complexity. We further validated the improved diversity of feature representations resulting from different self-attention heads, and our model design against other sparse attention mechanisms.
Current universal segmentation methods demonstrate strong capabilities in pixel-level image and video understanding. However, they lack reasoning abilities and cannot be controlled via text instructions. In contrast, large vision-language multimodal models exhibit powerful vision-based conversation and reasoning capabilities but lack pixel-level understanding and have difficulty accepting visual prompts for flexible user interaction. This paper proposes OMG-LLaVA, a new and elegant framework combining powerful pixel-level vision understanding with reasoning abilities. It can accept various visual and text prompts for flexible user interaction. Specifically, we use a universal segmentation method as the visual encoder, integrating image information, perception priors, and visual prompts into visual tokens provided to the LLM. The LLM is responsible for understanding the user's text instructions and providing text responses and pixel-level segmentation results based on the visual information. We propose perception prior embedding to better integrate perception priors with image features. OMG-LLaVA achieves image-level, object-level, and pixel-level reasoning and understanding in a single model, matching or surpassing the performance of specialized methods on multiple benchmarks. Rather than using LLM to connect each specialist, our work aims at end-to-end training on one encoder, one decoder, and one LLM. The code and model have been released for further research.
The Long-form Video Question-Answering task requires the comprehension and analysis of extended video content to respond accurately to questions by utilizing both temporal and contextual information. In this paper, we present MM-Screenplayer, an advanced video understanding system with multi-modal perception capabilities that can convert any video into textual screenplay representations. Unlike previous storytelling methods, we organize video content into scenes as the basic unit, rather than just visually continuous shots. Additionally, we developed a ``Look Back'' strategy to reassess and validate uncertain information, particularly targeting breakpoint mode. MM-Screenplayer achieved highest score in the CVPR'2024 LOng-form VidEo Understanding (LOVEU) Track 1 Challenge, with a global accuracy of 87.5% and a breakpoint accuracy of 68.8%.
Pixel-level Video Understanding in the Wild Challenge (PVUW) focus on complex video understanding. In this CVPR 2024 workshop, we add two new tracks, Complex Video Object Segmentation Track based on MOSE dataset and Motion Expression guided Video Segmentation track based on MeViS dataset. In the two new tracks, we provide additional videos and annotations that feature challenging elements, such as the disappearance and reappearance of objects, inconspicuous small objects, heavy occlusions, and crowded environments in MOSE. Moreover, we provide a new motion expression guided video segmentation dataset MeViS to study the natural language-guided video understanding in complex environments. These new videos, sentences, and annotations enable us to foster the development of a more comprehensive and robust pixel-level understanding of video scenes in complex environments and realistic scenarios. The MOSE challenge had 140 registered teams in total, 65 teams participated the validation phase and 12 teams made valid submissions in the final challenge phase. The MeViS challenge had 225 registered teams in total, 50 teams participated the validation phase and 5 teams made valid submissions in the final challenge phase.
Recent advancements in Multimodal Large Language Models (MLLMs) have extended their capabilities to video understanding. Yet, these models are often plagued by "hallucinations", where irrelevant or nonsensical content is generated, deviating from the actual video context. This work introduces VideoHallucer, the first comprehensive benchmark for hallucination detection in large video-language models (LVLMs). VideoHallucer categorizes hallucinations into two main types: intrinsic and extrinsic, offering further subcategories for detailed analysis, including object-relation, temporal, semantic detail, extrinsic factual, and extrinsic non-factual hallucinations. We adopt an adversarial binary VideoQA method for comprehensive evaluation, where pairs of basic and hallucinated questions are crafted strategically. By evaluating eleven LVLMs on VideoHallucer, we reveal that i) the majority of current models exhibit significant issues with hallucinations; ii) while scaling datasets and parameters improves models' ability to detect basic visual cues and counterfactuals, it provides limited benefit for detecting extrinsic factual hallucinations; iii) existing models are more adept at detecting facts than identifying hallucinations. As a byproduct, these analyses further instruct the development of our self-PEP framework, achieving an average of 5.38% improvement in hallucination resistance across all model architectures.
Speech understanding as an element of the more generic video understanding using audio-visual large language models (av-LLMs) is a crucial yet understudied aspect. This paper proposes video-SALMONN, a single end-to-end av-LLM for video processing, which can understand not only visual frame sequences, audio events and music, but speech as well. To obtain fine-grained temporal information required by speech understanding, while keeping efficient for other video elements, this paper proposes a novel multi-resolution causal Q-Former (MRC Q-Former) structure to connect pre-trained audio-visual encoders and the backbone large language model. Moreover, dedicated training approaches including the diversity loss and the unpaired audio-visual mixed training scheme are proposed to avoid frames or modality dominance. On the introduced speech-audio-visual evaluation benchmark, video-SALMONN achieves more than 25\% absolute accuracy improvements on the video-QA task and over 30\% absolute accuracy improvements on audio-visual QA tasks with human speech. In addition, video-SALMONN demonstrates remarkable video comprehension and reasoning abilities on tasks that are unprecedented by other av-LLMs. Our training code and model checkpoints are available at \texttt{\url{https://github.com/bytedance/SALMONN/}}.
Recent advancements in Large Language Models (LLMs) have expanded their capabilities to multimodal contexts, including comprehensive video understanding. However, processing extensive videos such as 24-hour CCTV footage or full-length films presents significant challenges due to the vast data and processing demands. Traditional methods, like extracting key frames or converting frames to text, often result in substantial information loss. To address these shortcomings, we develop OmAgent, efficiently stores and retrieves relevant video frames for specific queries, preserving the detailed content of videos. Additionally, it features an Divide-and-Conquer Loop capable of autonomous reasoning, dynamically invoking APIs and tools to enhance query processing and accuracy. This approach ensures robust video understanding, significantly reducing information loss. Experimental results affirm OmAgent's efficacy in handling various types of videos and complex tasks. Moreover, we have endowed it with greater autonomy and a robust tool-calling system, enabling it to accomplish even more intricate tasks.
A more robust and holistic language-video representation is the key to pushing video understanding forward. Despite the improvement in training strategies, the quality of the language-video dataset is less attention to. The current plain and simple text descriptions and the visual-only focus for the language-video tasks result in a limited capacity in real-world natural language video retrieval tasks where queries are much more complex. This paper introduces a method to automatically enhance video-language datasets, making them more modality and context-aware for more sophisticated representation learning needs, hence helping all downstream tasks. Our multifaceted video captioning method captures entities, actions, speech transcripts, aesthetics, and emotional cues, providing detailed and correlating information from the text side to the video side for training. We also develop an agent-like strategy using language models to generate high-quality, factual textual descriptions, reducing human intervention and enabling scalability. The method's effectiveness in improving language-video representation is evaluated through text-video retrieval using the MSR-VTT dataset and several multi-modal retrieval models.
Remote physiological measurement (RPM) is an essential tool for healthcare monitoring as it enables the measurement of physiological signs, e.g., heart rate, in a remote setting via physical wearables. Recently, with facial videos, we have seen rapid advancements in video-based RPMs. However, adopting facial videos for RPM in the clinical setting largely depends on the accuracy and robustness (work across patient populations). Fortunately, the capability of the state-of-the-art transformer architecture in general (natural) video understanding has resulted in marked improvements and has been translated to facial understanding, including RPM. However, existing RPM methods usually need RPM-specific modules, e.g., temporal difference convolution and handcrafted feature maps. Although these customized modules can increase accuracy, they are not demonstrated for their robustness across datasets. Further, due to their customization of the transformer architecture, they cannot use the advancements made in general video transformers (GVT). In this study, we interrogate the GVT architecture and empirically analyze how the training designs, i.e., data pre-processing and network configurations, affect the model performance applied to RPM. Based on the structure of video transformers, we propose to configure its spatiotemporal hierarchy to align with the dense temporal information needed in RPM for signal feature extraction. We define several practical guidelines and gradually adapt GVTs for RPM without introducing RPM-specific modules. Our experiments demonstrate favorable results to existing RPM-specific module counterparts. We conducted extensive experiments with five datasets using intra-dataset and cross-dataset settings. We highlight that the proposed guidelines GVT2RPM can be generalized to any video transformers and is robust to various datasets.