Spatio-temporal action detection (STAD) is an important fine-grained video understanding task. Current methods require box and label supervision for all action classes in advance. However, in real-world applications, it is very likely to come across new action classes not seen in training because the action category space is large and hard to enumerate. Also, the cost of data annotation and model training for new classes is extremely high for traditional methods, as we need to perform detailed box annotations and re-train the whole network from scratch. In this paper, we propose a new challenging setting by performing open-vocabulary STAD to better mimic the situation of action detection in an open world. Open-vocabulary spatio-temporal action detection (OV-STAD) requires training a model on a limited set of base classes with box and label supervision, which is expected to yield good generalization performance on novel action classes. For OV-STAD, we build two benchmarks based on the existing STAD datasets and propose a simple but effective method based on pretrained video-language models (VLM). To better adapt the holistic VLM for the fine-grained action detection task, we carefully fine-tune it on the localized video region-text pairs. This customized fine-tuning endows the VLM with better motion understanding, thus contributing to a more accurate alignment between video regions and texts. Local region feature and global video feature fusion before alignment is adopted to further improve the action detection performance by providing global context. Our method achieves a promising performance on novel classes.
Transformer-based long context generative models power emerging AI applications like hour-long video understanding and project-level coding agent. Deploying long context transformers (e.g., 100K to 10M tokens) is prohibitively expensive compared to short context (e.g., 4K tokens) model variants. Reducing the cost of long-context transformers is becoming a pressing research and engineering challenge starting from the year of 2024. This work describes a concurrent programming framework for quantitatively analyzing the efficiency challenges in serving multiple long-context requests under limited size of GPU high-bandwidth memory (HBM) regime. We give a detailed analysis of how all additional computational costs, compared to 4K context, trace back to \textit{one single source: the large size of the KV cache}. We use a 34B GPT-3.5 level model of 50K context on A100 NVLink as a running example, and describe how its large KV cache causes four types of deployment challenges: (1) prefilling long inputs takes much longer compute time and GPU memory than short inputs; (2) after prefilling, the large KV cache residing on the GPU HBM substantially restricts the number of concurrent users being served; (3) during decoding, repeatedly reading the KV cache from HBM to SM largely increases latency; (4) when KV cache memory overflows, swapping it from HBM to DDR causes significant context switching latency. We use this framework to analyze existing works and identify possibilities of combining them to build end-to-end systems. Overall, this work offers a foundational framework for analyzing long context transformer deployment and identifies directions towards reducing the inference cost of 1M context to be as cheap as 4K.
Current architectures for video understanding mainly build upon 3D convolutional blocks or 2D convolutions with additional operations for temporal modeling. However, these methods all regard the temporal axis as a separate dimension of the video sequence, which requires large computation and memory budgets and thus limits their usage on mobile devices. In this paper, we propose to squeeze the time axis of a video sequence into the channel dimension and present a lightweight video recognition network, term as \textit{SqueezeTime}, for mobile video understanding. To enhance the temporal modeling capability of the proposed network, we design a Channel-Time Learning (CTL) Block to capture temporal dynamics of the sequence. This module has two complementary branches, in which one branch is for temporal importance learning and another branch with temporal position restoring capability is to enhance inter-temporal object modeling ability. The proposed SqueezeTime is much lightweight and fast with high accuracies for mobile video understanding. Extensive experiments on various video recognition and action detection benchmarks, i.e., Kinetics400, Kinetics600, HMDB51, AVA2.1 and THUMOS14, demonstrate the superiority of our model. For example, our SqueezeTime achieves $+1.2\%$ accuracy and $+80\%$ GPU throughput gain on Kinetics400 than prior methods. Codes are publicly available at https://github.com/xinghaochen/SqueezeTime and https://github.com/mindspore-lab/models/tree/master/research/huawei-noah/SqueezeTime.
Current datasets for long-form video understanding often fall short of providing genuine long-form comprehension challenges, as many tasks derived from these datasets can be successfully tackled by analyzing just one or a few random frames from a video. To address this issue, we present a novel dataset and benchmark, CinePile, specifically designed for authentic long-form video understanding. This paper details our innovative approach for creating a question-answer dataset, utilizing advanced LLMs with human-in-the-loop and building upon human-generated raw data. Our comprehensive dataset comprises 305,000 multiple-choice questions (MCQs), covering various visual and multimodal aspects, including temporal comprehension, understanding human-object interactions, and reasoning about events or actions within a scene. Additionally, we fine-tuned open-source Video-LLMs on the training split and evaluated both open-source and proprietary video-centric LLMs on the test split of our dataset. The findings indicate that although current models underperform compared to humans, fine-tuning these models can lead to significant improvements in their performance.
Despite the significant progress of fully-supervised video captioning, zero-shot methods remain much less explored. In this paper, we propose a novel zero-shot video captioning framework named Retrieval-Enhanced Test-Time Adaptation (RETTA), which takes advantage of existing pretrained large-scale vision and language models to directly generate captions with test-time adaptation. Specifically, we bridge video and text using four key models: a general video-text retrieval model XCLIP, a general image-text matching model CLIP, a text alignment model AnglE, and a text generation model GPT-2, due to their source-code availability. The main challenge is how to enable the text generation model to be sufficiently aware of the content in a given video so as to generate corresponding captions. To address this problem, we propose using learnable tokens as a communication medium among these four frozen models GPT-2, XCLIP, CLIP, and AnglE. Different from the conventional way that trains these tokens with training data, we propose to learn these tokens with soft targets of the inference data under several carefully crafted loss functions, which enable the tokens to absorb video information catered for GPT-2. This procedure can be efficiently done in just a few iterations (we use 16 iterations in the experiments) and does not require ground truth data. Extensive experimental results on three widely used datasets, MSR-VTT, MSVD, and VATEX, show absolute 5.1%-32.4% improvements in terms of the main metric CIDEr compared to several state-of-the-art zero-shot video captioning methods.
In this paper, we show that transferring knowledge from other domains of video understanding combined with large-scale learning can improve robustness of Video Object Segmentation (VOS) under complex circumstances. Namely, we focus on integrating scene global motion knowledge to improve large-scale semi-supervised Video Object Segmentation. Prior works on VOS mostly rely on direct comparison of semantic and contextual features to perform dense matching between current and past frames, passing over actual motion structure. On the other hand, Optical Flow Estimation task aims to approximate the scene motion field, exposing global motion patterns which are typically undiscoverable during all pairs similarity search. We present WarpFormer, an architecture for semi-supervised Video Object Segmentation that exploits existing knowledge in motion understanding to conduct smoother propagation and more accurate matching. Our framework employs a generic pretrained Optical Flow Estimation network whose prediction is used to warp both past frames and instance segmentation masks to the current frame domain. Consequently, warped segmentation masks are refined and fused together aiming to inpaint occluded regions and eliminate artifacts caused by flow field imperfects. Additionally, we employ novel large-scale MOSE 2023 dataset to train model on various complex scenarios. Our method demonstrates strong performance on DAVIS 2016/2017 validation (93.0% and 85.9%), DAVIS 2017 test-dev (80.6%) and YouTube-VOS 2019 validation (83.8%) that is competitive with alternative state-of-the-art methods while using much simpler memory mechanism and instance understanding logic.
Action recognition is a key technology in building interactive metaverses. With the rapid development of deep learning, methods in action recognition have also achieved great advancement. Researchers design and implement the backbones referring to multiple standpoints, which leads to the diversity of methods and encountering new challenges. This paper reviews several action recognition methods based on deep neural networks. We introduce these methods in three parts: 1) Two-Streams networks and their variants, which, specifically in this paper, use RGB video frame and optical flow modality as input; 2) 3D convolutional networks, which make efforts in taking advantage of RGB modality directly while extracting different motion information is no longer necessary; 3) Transformer-based methods, which introduce the model from natural language processing into computer vision and video understanding. We offer objective sights in this review and hopefully provide a reference for future research.
Video Foundation Models (ViFMs) aim to learn a general-purpose representation for various video understanding tasks. Leveraging large-scale datasets and powerful models, ViFMs achieve this by capturing robust and generic features from video data. This survey analyzes over 200 video foundational models, offering a comprehensive overview of benchmarks and evaluation metrics across 14 distinct video tasks categorized into 3 main categories. Additionally, we offer an in-depth performance analysis of these models for the 6 most common video tasks. We categorize ViFMs into three categories: 1) Image-based ViFMs, which adapt existing image models for video tasks, 2) Video-Based ViFMs, which utilize video-specific encoding methods, and 3) Universal Foundational Models (UFMs), which combine multiple modalities (image, video, audio, and text etc.) within a single framework. By comparing the performance of various ViFMs on different tasks, this survey offers valuable insights into their strengths and weaknesses, guiding future advancements in video understanding. Our analysis surprisingly reveals that image-based foundation models consistently outperform video-based models on most video understanding tasks. Additionally, UFMs, which leverage diverse modalities, demonstrate superior performance on video tasks. We share the comprehensive list of ViFMs studied in this work at: \url{https://github.com/NeeluMadan/ViFM_Survey.git}
State Space Model (SSM) is a mathematical model used to describe and analyze the behavior of dynamic systems. This model has witnessed numerous applications in several fields, including control theory, signal processing, economics and machine learning. In the field of deep learning, state space models are used to process sequence data, such as time series analysis, natural language processing (NLP) and video understanding. By mapping sequence data to state space, long-term dependencies in the data can be better captured. In particular, modern SSMs have shown strong representational capabilities in NLP, especially in long sequence modeling, while maintaining linear time complexity. Notably, based on the latest state-space models, Mamba merges time-varying parameters into SSMs and formulates a hardware-aware algorithm for efficient training and inference. Given its impressive efficiency and strong long-range dependency modeling capability, Mamba is expected to become a new AI architecture that may outperform Transformer. Recently, a number of works have attempted to study the potential of Mamba in various fields, such as general vision, multi-modal, medical image analysis and remote sensing image analysis, by extending Mamba from natural language domain to visual domain. To fully understand Mamba in the visual domain, we conduct a comprehensive survey and present a taxonomy study. This survey focuses on Mamba's application to a variety of visual tasks and data types, and discusses its predecessors, recent advances and far-reaching impact on a wide range of domains. Since Mamba is now on an upward trend, please actively notice us if you have new findings, and new progress on Mamba will be included in this survey in a timely manner and updated on the Mamba project at https://github.com/lx6c78/Vision-Mamba-A-Comprehensive-Survey-and-Taxonomy.
Multimodal information, together with our knowledge, help us to understand the complex and dynamic world. Large language models (LLM) and large multimodal models (LMM), however, still struggle to emulate this capability. In this paper, we present WorldQA, a video understanding dataset designed to push the boundaries of multimodal world models with three appealing properties: (1) Multimodal Inputs: The dataset comprises 1007 question-answer pairs and 303 videos, necessitating the analysis of both auditory and visual data for successful interpretation. (2) World Knowledge: We identify five essential types of world knowledge for question formulation. This approach challenges models to extend their capabilities beyond mere perception. (3) Long-Chain Reasoning: Our dataset introduces an average reasoning step of 4.45, notably surpassing other videoQA datasets. Furthermore, we introduce WorldRetriever, an agent designed to synthesize expert knowledge into a coherent reasoning chain, thereby facilitating accurate responses to WorldQA queries. Extensive evaluations of 13 prominent LLMs and LMMs reveal that WorldRetriever, although being the most effective model, achieved only 70% of humanlevel performance in multiple-choice questions. This finding highlights the necessity for further advancement in the reasoning and comprehension abilities of models. Our experiments also yield several key insights. For instance, while humans tend to perform better with increased frames, current LMMs, including WorldRetriever, show diminished performance under similar conditions. We hope that WorldQA,our methodology, and these insights could contribute to the future development of multimodal world models.
Recent advancements in Large Language Models (LLMs) have led to the development of Video Large Multi-modal Models (Video-LMMs) that can handle a wide range of video understanding tasks. These models have the potential to be deployed in real-world applications such as robotics, AI assistants, medical surgery, and autonomous vehicles. The widespread adoption of Video-LMMs in our daily lives underscores the importance of ensuring and evaluating their robust performance in mirroring human-like reasoning and interaction capabilities in complex, real-world contexts. However, existing benchmarks for Video-LMMs primarily focus on general video comprehension abilities and neglect assessing their reasoning capabilities over complex videos in the real-world context, and robustness of these models through the lens of user prompts as text queries. In this paper, we present the Complex Video Reasoning and Robustness Evaluation Suite (CVRR-ES), a novel benchmark that comprehensively assesses the performance of Video-LMMs across 11 diverse real-world video dimensions. We evaluate 9 recent models, including both open-source and closed-source variants, and find that most of the Video-LMMs, especially open-source ones, struggle with robustness and reasoning when dealing with complex videos. Based on our analysis, we develop a training-free Dual-Step Contextual Prompting (DSCP) technique to enhance the performance of existing Video-LMMs. Our findings provide valuable insights for building the next generation of human-centric AI systems with advanced robustness and reasoning capabilities. Our dataset and code are publicly available at: https://mbzuai-oryx.github.io/CVRR-Evaluation-Suite/.
Recently, integrating video foundation models and large language models to build a video understanding system can overcome the limitations of specific pre-defined vision tasks. Yet, existing methods either employ complex spatial-temporal modules or rely heavily on additional perception models to extract temporal features for video understanding, and they only perform well on short videos. For long videos, the computational complexity and memory costs associated with long-term temporal connections are significantly increased, posing additional challenges.Taking advantage of the Atkinson-Shiffrin memory model, with tokens in Transformers being employed as the carriers of memory in combination with our specially designed memory mechanism, we propose MovieChat to overcome these challenges. We lift pre-trained multi-modal large language models for understanding long videos without incorporating additional trainable temporal modules, employing a zero-shot approach. MovieChat achieves state-of-the-art performance in long video understanding, along with the released MovieChat-1K benchmark with 1K long video, 2K temporal grounding labels, and 14K manual annotations for validation of the effectiveness of our method. The code along with the dataset can be accessed via the following https://github.com/rese1f/MovieChat.
In Video-based Facial Expression Recognition (V-FER), models are typically trained on closed-set datasets with a fixed number of known classes. However, these V-FER models cannot deal with unknown classes that are prevalent in real-world scenarios. In this paper, we introduce a challenging Open-set Video-based Facial Expression Recognition (OV-FER) task, aiming at identifying not only known classes but also new, unknown human facial expressions not encountered during training. While existing approaches address open-set recognition by leveraging large-scale vision-language models like CLIP to identify unseen classes, we argue that these methods may not adequately capture the nuanced and subtle human expression patterns required by the OV-FER task. To address this limitation, we propose a novel Human Expression-Sensitive Prompting (HESP) mechanism to significantly enhance CLIP's ability to model video-based facial expression details effectively, thereby presenting a new CLIP-based OV-FER approach. Our proposed HESP comprises three components: 1) a textual prompting module with learnable prompt representations to complement the original CLIP textual prompts and enhance the textual representations of both known and unknown emotions, 2) a visual prompting module that encodes temporal emotional information from video frames using expression-sensitive attention, equipping CLIP with a new visual modeling ability to extract emotion-rich information, 3) a delicately designed open-set multi-task learning scheme that facilitates prompt learning and encourages interactions between the textual and visual prompting modules. Extensive experiments conducted on four OV-FER task settings demonstrate that HESP can significantly boost CLIP's performance (a relative improvement of 17.93% on AUROC and 106.18% on OSCR) and outperform other state-of-the-art open-set video understanding methods by a large margin.
Vision-language pre-training has significantly elevated performance across a wide range of image-language applications. Yet, the pre-training process for video-related tasks demands exceptionally large computational and data resources, which hinders the progress of video-language models. This paper investigates a straight-forward, highly efficient, and resource-light approach to adapting an existing image-language pre-trained model for dense video understanding. Our preliminary experiments reveal that directly fine-tuning pre-trained image-language models with multiple frames as inputs on video datasets leads to performance saturation or even a drop. Our further investigation reveals that it is largely attributed to the bias of learned high-norm visual features. Motivated by this finding, we propose a simple but effective pooling strategy to smooth the feature distribution along the temporal dimension and thus reduce the dominant impacts from the extreme features. The new model is termed Pooling LLaVA, or PLLaVA in short. PLLaVA achieves new state-of-the-art performance on modern benchmark datasets for both video question-answer and captioning tasks. Notably, on the recent popular VideoChatGPT benchmark, PLLaVA achieves a score of 3.48 out of 5 on average of five evaluated dimensions, exceeding the previous SOTA results from GPT4V (IG-VLM) by 9%. On the latest multi-choice benchmark MVBench, PLLaVA achieves 58.1% accuracy on average across 20 sub-tasks, 14.5% higher than GPT4V (IG-VLM). Code is available at https://pllava.github.io/
We describe a protocol to study text-to-video retrieval training with unlabeled videos, where we assume (i) no access to labels for any videos, i.e., no access to the set of ground-truth captions, but (ii) access to labeled images in the form of text. Using image expert models is a realistic scenario given that annotating images is cheaper therefore scalable, in contrast to expensive video labeling schemes. Recently, zero-shot image experts such as CLIP have established a new strong baseline for video understanding tasks. In this paper, we make use of this progress and instantiate the image experts from two types of models: a text-to-image retrieval model to provide an initial backbone, and image captioning models to provide supervision signal into unlabeled videos. We show that automatically labeling video frames with image captioning allows text-to-video retrieval training. This process adapts the features to the target domain at no manual annotation cost, consequently outperforming the strong zero-shot CLIP baseline. During training, we sample captions from multiple video frames that best match the visual content, and perform a temporal pooling over frame representations by scoring frames according to their relevance to each caption. We conduct extensive ablations to provide insights and demonstrate the effectiveness of this simple framework by outperforming the CLIP zero-shot baselines on text-to-video retrieval on three standard datasets, namely ActivityNet, MSR-VTT, and MSVD.
This paper offers an insightful examination of how currently top-trending AI technologies, i.e., generative artificial intelligence (Generative AI) and large language models (LLMs), are reshaping the field of video technology, including video generation, understanding, and streaming. It highlights the innovative use of these technologies in producing highly realistic videos, a significant leap in bridging the gap between real-world dynamics and digital creation. The study also delves into the advanced capabilities of LLMs in video understanding, demonstrating their effectiveness in extracting meaningful information from visual content, thereby enhancing our interaction with videos. In the realm of video streaming, the paper discusses how LLMs contribute to more efficient and user-centric streaming experiences, adapting content delivery to individual viewer preferences. This comprehensive review navigates through the current achievements, ongoing challenges, and future possibilities of applying Generative AI and LLMs to video-related tasks, underscoring the immense potential these technologies hold for advancing the field of video technology related to multimedia, networking, and AI communities.
The task of spatiotemporal action localization in chaotic scenes is a challenging task toward advanced video understanding. Paving the way with high-quality video feature extraction and enhancing the precision of detector-predicted anchors can effectively improve model performance. To this end, we propose a high-performance dual-stream spatiotemporal feature extraction network SFMViT with an anchor pruning strategy. The backbone of our SFMViT is composed of ViT and SlowFast with prior knowledge of spatiotemporal action localization, which fully utilizes ViT's excellent global feature extraction capabilities and SlowFast's spatiotemporal sequence modeling capabilities. Secondly, we introduce the confidence maximum heap to prune the anchors detected in each frame of the picture to filter out the effective anchors. These designs enable our SFMViT to achieve a mAP of 26.62% in the Chaotic World dataset, far exceeding existing models. Code is available at https://github.com/jfightyr/SlowFast-Meet-ViT.
Video anomaly detection (VAD) is a challenging task aiming to recognize anomalies in video frames, and existing large-scale VAD researches primarily focus on road traffic and human activity scenes. In industrial scenes, there are often a variety of unpredictable anomalies, and the VAD method can play a significant role in these scenarios. However, there is a lack of applicable datasets and methods specifically tailored for industrial production scenarios due to concerns regarding privacy and security. To bridge this gap, we propose a new dataset, IPAD, specifically designed for VAD in industrial scenarios. The industrial processes in our dataset are chosen through on-site factory research and discussions with engineers. This dataset covers 16 different industrial devices and contains over 6 hours of both synthetic and real-world video footage. Moreover, we annotate the key feature of the industrial process, ie, periodicity. Based on the proposed dataset, we introduce a period memory module and a sliding window inspection mechanism to effectively investigate the periodic information in a basic reconstruction model. Our framework leverages LoRA adapter to explore the effective migration of pretrained models, which are initially trained using synthetic data, into real-world scenarios. Our proposed dataset and method will fill the gap in the field of industrial video anomaly detection and drive the process of video understanding tasks as well as smart factory deployment.
Covering from Image LLMs to the more complex Video LLMs, the Multimodal Large Language Models (MLLMs) have demonstrated profound capabilities in comprehending cross-modal information as numerous studies have illustrated. Previous methods delve into designing comprehensive Video LLMs through integrating video foundation models with primitive LLMs. Despite its effectiveness, such paradigm renders Video LLM's structure verbose and typically requires substantial video data for pre-training. Crucially, it neglects leveraging the foundational contributions of ready-made Image LLMs. In this paper, we introduce RED-VILLM, a Resource-Efficient Development pipeline which builds robust Video LLMs through leveraging the prior knowledge of Image LLMs. Specifically, since a video is naturally a combination of images along the temporal dimension, we devise a temporal adaptation plug-and-play structure, endowing the backbone Image LLM with the capability to grasp temporal information. Moreover, through applying this pipeline, we achieve the first Video LLM within the Chinese-speaking community. Extensive experiments demonstrate that Video LLMs developed through our approach surpass conventional Video LLMs, requiring minimal instructional data and training resources. Our approach highlights the potential for a more cost-effective and scalable advancement in multimodal models.
We propose a novel framework for video understanding, called Temporally Contextualized CLIP (TC-CLIP), which leverages essential temporal information through global interactions in a spatio-temporal domain within a video. To be specific, we introduce Temporal Contextualization (TC), a layer-wise temporal information infusion mechanism for videos, which 1) extracts core information from each frame, 2) connects relevant information across frames for the summarization into context tokens, and 3) leverages the context tokens for feature encoding. Furthermore, the Video-conditional Prompting (VP) module processes context tokens to generate informative prompts in the text modality. Extensive experiments in zero-shot, few-shot, base-to-novel, and fully-supervised action recognition validate the effectiveness of our model. Ablation studies for TC and VP support our design choices. Our project page with the source code is available at https://github.com/naver-ai/tc-clip