Surgical video understanding is pivotal for enabling automated intraoperative decision-making, skill assessment, and postoperative quality improvement. However, progress in developing surgical video foundation models (FMs) remains hindered by the scarcity of large-scale, diverse datasets for pretraining and systematic evaluation. In this paper, we introduce \textbf{SurgBench}, a unified surgical video benchmarking framework comprising a pretraining dataset, \textbf{SurgBench-P}, and an evaluation benchmark, \textbf{SurgBench-E}. SurgBench offers extensive coverage of diverse surgical scenarios, with SurgBench-P encompassing 53 million frames across 22 surgical procedures and 11 specialties, and SurgBench-E providing robust evaluation across six categories (phase classification, camera motion, tool recognition, disease diagnosis, action classification, and organ detection) spanning 72 fine-grained tasks. Extensive experiments reveal that existing video FMs struggle to generalize across varied surgical video analysis tasks, whereas pretraining on SurgBench-P yields substantial performance improvements and superior cross-domain generalization to unseen procedures and modalities. Our dataset and code are available upon request.
Composed Video Retrieval (CoVR) retrieves a target video given a query video and a modification text describing the intended change. Existing CoVR benchmarks emphasize appearance shifts or coarse event changes and therefore do not test the ability to capture subtle, fast-paced temporal differences. We introduce TF-CoVR, the first large-scale benchmark dedicated to temporally fine-grained CoVR. TF-CoVR focuses on gymnastics and diving and provides 180K triplets drawn from FineGym and FineDiving. Previous CoVR benchmarks focusing on temporal aspect, link each query to a single target segment taken from the same video, limiting practical usefulness. In TF-CoVR, we instead construct each <query, modification> pair by prompting an LLM with the label differences between clips drawn from different videos; every pair is thus associated with multiple valid target videos (3.9 on average), reflecting real-world tasks such as sports-highlight generation. To model these temporal dynamics we propose TF-CoVR-Base, a concise two-stage training framework: (i) pre-train a video encoder on fine-grained action classification to obtain temporally discriminative embeddings; (ii) align the composed query with candidate videos using contrastive learning. We conduct the first comprehensive study of image, video, and general multimodal embedding (GME) models on temporally fine-grained composed retrieval in both zero-shot and fine-tuning regimes. On TF-CoVR, TF-CoVR-Base improves zero-shot mAP@50 from 5.92 (LanguageBind) to 7.51, and after fine-tuning raises the state-of-the-art from 19.83 to 25.82.
Traditional approaches in unsupervised or self supervised learning for skeleton-based action classification have concentrated predominantly on the dynamic aspects of skeletal sequences. Yet, the intricate interaction between the moving and static elements of the skeleton presents a rarely tapped discriminative potential for action classification. This paper introduces a novel measurement, referred to as spatial-temporal joint density (STJD), to quantify such interaction. Tracking the evolution of this density throughout an action can effectively identify a subset of discriminative moving and/or static joints termed "prime joints" to steer self-supervised learning. A new contrastive learning strategy named STJD-CL is proposed to align the representation of a skeleton sequence with that of its prime joints while simultaneously contrasting the representations of prime and nonprime joints. In addition, a method called STJD-MP is developed by integrating it with a reconstruction-based framework for more effective learning. Experimental evaluations on the NTU RGB+D 60, NTU RGB+D 120, and PKUMMD datasets in various downstream tasks demonstrate that the proposed STJD-CL and STJD-MP improved performance, particularly by 3.5 and 3.6 percentage points over the state-of-the-art contrastive methods on the NTU RGB+D 120 dataset using X-sub and X-set evaluations, respectively.
The integration of artificial intelligence in sports analytics has transformed soccer video understanding, enabling real-time, automated insights into complex game dynamics. Traditional approaches rely on isolated data streams, limiting their effectiveness in capturing the full context of a match. To address this, we introduce SoccerChat, a multimodal conversational AI framework that integrates visual and textual data for enhanced soccer video comprehension. Leveraging the extensive SoccerNet dataset, enriched with jersey color annotations and automatic speech recognition (ASR) transcripts, SoccerChat is fine-tuned on a structured video instruction dataset to facilitate accurate game understanding, event classification, and referee decision making. We benchmark SoccerChat on action classification and referee decision-making tasks, demonstrating its performance in general soccer event comprehension while maintaining competitive accuracy in referee decision making. Our findings highlight the importance of multimodal integration in advancing soccer analytics, paving the way for more interactive and explainable AI-driven sports analysis. https://github.com/simula/SoccerChat
Machine learning and computer vision methods have a major impact on the study of natural animal behavior, as they enable the (semi-)automatic analysis of vast amounts of video data. Mice are the standard mammalian model system in most research fields, but the datasets available today to refine such methods focus either on simple or social behaviors. In this work, we present a video dataset of individual mice solving complex mechanical puzzles, so-called lockboxes. The more than 110 hours of total playtime show their behavior recorded from three different perspectives. As a benchmark for frame-level action classification methods, we provide human-annotated labels for all videos of two different mice, that equal 13% of our dataset. Our keypoint (pose) tracking-based action classification framework illustrates the challenges of automated labeling of fine-grained behaviors, such as the manipulation of objects. We hope that our work will help accelerate the advancement of automated action and behavior classification in the computational neuroscience community. Our dataset is publicly available at https://doi.org/10.14279/depositonce-23850
Vision Language Models (VLMs) have demonstrated strong performance in multi-modal tasks by effectively aligning visual and textual representations. However, most video understanding VLM research has been domain-agnostic, leaving the understanding of their transfer learning capability to specialized domains under-explored. In this work, we address this by exploring the adaptability of open-source VLMs to specific domains, and focusing on soccer as an initial case study. Our approach uses large-scale soccer datasets and LLM to create instruction-following data, and use them to iteratively fine-tune the general-domain VLM in a curriculum learning fashion (first teaching the model key soccer concepts to then question answering tasks). The final adapted model, trained using a curated dataset of 20k video clips, exhibits significant improvement in soccer-specific tasks compared to the base model, with a 37.5% relative improvement for the visual question-answering task and an accuracy improvement from 11.8% to 63.5% for the downstream soccer action classification task.
We propose Cross-Attention in Audio, Space, and Time (CA^2ST), a transformer-based method for holistic video recognition. Recognizing actions in videos requires both spatial and temporal understanding, yet most existing models lack a balanced spatio-temporal understanding of videos. To address this, we propose a novel two-stream architecture, called Cross-Attention in Space and Time (CAST), using only RGB input. In each layer of CAST, Bottleneck Cross-Attention (B-CA) enables spatial and temporal experts to exchange information and make synergistic predictions. For holistic video understanding, we extend CAST by integrating an audio expert, forming Cross-Attention in Visual and Audio (CAVA). We validate the CAST on benchmarks with different characteristics, EPIC-KITCHENS-100, Something-Something-V2, and Kinetics-400, consistently showing balanced performance. We also validate the CAVA on audio-visual action recognition benchmarks, including UCF-101, VGG-Sound, KineticsSound, and EPIC-SOUNDS. With a favorable performance of CAVA across these datasets, we demonstrate the effective information exchange among multiple experts within the B-CA module. In summary, CA^2ST combines CAST and CAVA by employing spatial, temporal, and audio experts through cross-attention, achieving balanced and holistic video understanding.
Human action recognition in low-light environments is crucial for various real-world applications. However, the existing approaches overlook the full utilization of brightness information throughout the training phase, leading to suboptimal performance. To address this limitation, we propose OwlSight, a biomimetic-inspired framework with whole-stage illumination enhancement to interact with action classification for accurate dark video human action recognition. Specifically, OwlSight incorporates a Time-Consistency Module (TCM) to capture shallow spatiotemporal features meanwhile maintaining temporal coherence, which are then processed by a Luminance Adaptation Module (LAM) to dynamically adjust the brightness based on the input luminance distribution. Furthermore, a Reflect Augmentation Module (RAM) is presented to maximize illumination utilization and simultaneously enhance action recognition via two interactive paths. Additionally, we build Dark-101, a large-scale dataset comprising 18,310 dark videos across 101 action categories, significantly surpassing existing datasets (e.g., ARID1.5 and Dark-48) in scale and diversity. Extensive experiments demonstrate that the proposed OwlSight achieves state-of-the-art performance across four low-light action recognition benchmarks. Notably, it outperforms previous best approaches by 5.36% on ARID1.5 and 1.72% on Dark-101, highlighting its effectiveness in challenging dark environments.
Popular video training methods mainly operate on a fixed number of tokens sampled from a predetermined spatiotemporal grid, resulting in sub-optimal accuracy-computation trade-offs due to inherent video redundancy. They also lack adaptability to varying computational budgets for downstream tasks, hindering applications of the most competitive model in real-world scenes. We thus propose a new test setting, Token Optimization, for maximized input information across budgets, which optimizes the size-limited set of input tokens through token selection from more suitably sampled videos. To this end, we propose a novel augmentation tool termed Flux. By making the sampling grid flexible and leveraging token selection, it is easily adopted in most popular video training frameworks, boosting model robustness with nearly no additional cost. We integrate Flux in large-scale video pre-training, and the resulting FluxViT establishes new state-of-the-art results across extensive tasks at standard costs. Notably, with 1/4 tokens only, it can still match the performance of previous state-of-the-art models with Token Optimization, yielding nearly 90\% savings. All models and data are available at https://github.com/OpenGVLab/FluxViT.
The way of living of many individuals around the world endures because of mental and physical disability associated with the movement of limbs. The usage of the assistive technology and systems will enhance the quality of affected people. In this situation, you can pave the way for a solution by transforming the movement of physical activities into computer-assisted applications. Surface Electromyography (sEMG) introduced the non-intervention procedure that can transform physical activities into signals for classification purposes and then practice it in applications. In this study, we suggest a scheme based on machine learning for the identification of 20 physical movements. This scheme follows up on the distinct characteristics from numerous signatures that include time-domain features, frequency-domain features, and inter-channel statistics of an sEMG signal. Afterward, we performed a thorough comparative examination of the k-NN and SVM classifier by considering the group of features for multiple normal and aggressive activities. The impact of different arrangements of dimensionalities has been recorded as well. Eventually, the SVM classifier gives 100% accuracy for 10 normal actions whereas 1-NN for a subgroup of features achieves 98.91% accuracy for 10 aggressive actions respectively. Additionally, we combine both SVM and 1-NN to propose a hybrid approach to classify 20 physical actions. The hybrid classifier gives an accuracy of 98.97% respectively. These recommendations are valuable for algorithm designers to select the finest approach by considering the resources available for the execution of an algorithm.
This paper introduces ViNet-S, a 36MB model based on the ViNet architecture with a U-Net design, featuring a lightweight decoder that significantly reduces model size and parameters without compromising performance. Additionally, ViNet-A (148MB) incorporates spatio-temporal action localization (STAL) features, differing from traditional video saliency models that use action classification backbones. Our studies show that an ensemble of ViNet-S and ViNet-A, by averaging predicted saliency maps, achieves state-of-the-art performance on three visual-only and six audio-visual saliency datasets, outperforming transformer-based models in both parameter efficiency and real-time performance, with ViNet-S reaching over 1000fps.
In competitive combat sports like boxing, analyzing a boxers's performance statics is crucial for evaluating the quantity and variety of punches delivered during bouts. These statistics provide valuable data and feedback, which are routinely used for coaching and performance enhancement. We introduce BoxMAC, a real-world boxing dataset featuring 15 professional boxers and encompassing 13 distinct action labels. Comprising over 60,000 frames, our dataset has been meticulously annotated for multiple actions per frame with inputs from a boxing coach. Since two boxers can execute different punches within a single timestamp, this problem falls under the domain of multi-label action classification. We propose a novel architecture for jointly recognizing multiple actions in both individual images and videos. We investigate baselines using deep neural network architectures to address both tasks. We believe that BoxMAC will enable researchers and practitioners to develop and evaluate more efficient models for performance analysis. With its realistic and diverse nature, BoxMAC can serve as a valuable resource for the advancement of boxing as a sport
Classifying fine-grained actions in fast-paced, close-combat sports such as fencing and boxing presents unique challenges due to the complexity, speed, and nuance of movements. Traditional methods reliant on pose estimation or fancy sensor data often struggle to capture these dynamics accurately. We introduce FACTS, a novel transformer-based approach for fine-grained action recognition that processes raw video data directly, eliminating the need for pose estimation and the use of cumbersome body markers and sensors. FACTS achieves state-of-the-art performance, with 90% accuracy on fencing actions and 83.25% on boxing actions. Additionally, we present a new publicly available dataset featuring 8 detailed fencing actions, addressing critical gaps in sports analytics resources. Our findings enhance training, performance analysis, and spectator engagement, setting a new benchmark for action classification in tactical sports.
Scaling has not yet been convincingly demonstrated for pure self-supervised learning from video. However, prior work has focused evaluations on semantic-related tasks $\unicode{x2013}$ action classification, ImageNet classification, etc. In this paper we focus on evaluating self-supervised learning on non-semantic vision tasks that are more spatial (3D) and temporal (+1D = 4D), such as camera pose estimation, point and object tracking, and depth estimation. We show that by learning from very large video datasets, masked auto-encoding (MAE) with transformer video models actually scales, consistently improving performance on these 4D tasks, as model size increases from 20M all the way to the largest by far reported self-supervised video model $\unicode{x2013}$ 22B parameters. Rigorous apples-to-apples comparison with many recent image and video models demonstrates the benefits of scaling 4D representations.
Existing skeleton-based human action classification models rely on well-trimmed action-specific skeleton videos for both training and testing, precluding their scalability to real-world applications where untrimmed videos exhibiting concatenated actions are predominant. To overcome this limitation, recently introduced skeleton action segmentation models involve un-trimmed skeleton videos into end-to-end training. The model is optimized to provide frame-wise predictions for any length of testing videos, simultaneously realizing action localization and classification. Yet, achieving such an improvement im-poses frame-wise annotated skeleton videos, which remains time-consuming in practice. This paper features a novel framework for skeleton-based action segmentation trained on short trimmed skeleton videos, but that can run on longer un-trimmed videos. The approach is implemented in three steps: Stitch, Contrast, and Segment. First, Stitch proposes a tem-poral skeleton stitching scheme that treats trimmed skeleton videos as elementary human motions that compose a semantic space and can be sampled to generate multi-action stitched se-quences. Contrast learns contrastive representations from stitched sequences with a novel discrimination pretext task that enables a skeleton encoder to learn meaningful action-temporal contexts to improve action segmentation. Finally, Segment relates the proposed method to action segmentation by learning a segmentation layer while handling particular da-ta availability. Experiments involve a trimmed source dataset and an untrimmed target dataset in an adaptation formulation for real-world skeleton-based human action segmentation to evaluate the effectiveness of the proposed method.
Temporal action localization (TAL) involves dual tasks to classify and localize actions within untrimmed videos. However, the two tasks often have conflicting requirements for features. Existing methods typically employ separate heads for classification and localization tasks but share the same input feature, leading to suboptimal performance. To address this issue, we propose a novel TAL method with Cross Layer Task Decoupling and Refinement (CLTDR). Based on the feature pyramid of video, CLTDR strategy integrates semantically strong features from higher pyramid layers and detailed boundary-aware boundary features from lower pyramid layers to effectively disentangle the action classification and localization tasks. Moreover, the multiple features from cross layers are also employed to refine and align the disentangled classification and regression results. At last, a lightweight Gated Multi-Granularity (GMG) module is proposed to comprehensively extract and aggregate video features at instant, local, and global temporal granularities. Benefiting from the CLTDR and GMG modules, our method achieves state-of-the-art performance on five challenging benchmarks: THUMOS14, MultiTHUMOS, EPIC-KITCHENS-100, ActivityNet-1.3, and HACS. Our code and pre-trained models are publicly available at: https://github.com/LiQiang0307/CLTDR-GMG.
Channel State Information (CSI) is the cornerstone in both wireless communication and sensing systems. In wireless communication systems, CSI provides essential insights into channel conditions, enabling system optimizations like channel compensation and dynamic resource allocation. However, the high computational complexity of CSI estimation algorithms necessitates the development of fast deep learning methods for CSI prediction. In wireless sensing systems, CSI can be leveraged to infer environmental changes, facilitating various functions, including gesture recognition and people identification. Deep learning methods have demonstrated significant advantages over model-based approaches in these fine-grained CSI classification tasks, particularly when classes vary across different scenarios. However, a major challenge in training deep learning networks for wireless systems is the limited availability of data, further complicated by the diverse formats of many public datasets, which hinder integration. Additionally, collecting CSI data can be resource-intensive, requiring considerable time and manpower. To address these challenges, we propose CSI-BERT2 for CSI prediction and classification tasks, effectively utilizing limited data through a pre-training and fine-tuning approach. Building on CSI-BERT1, we enhance the model architecture by introducing an Adaptive Re-Weighting Layer (ARL) and a Multi-Layer Perceptron (MLP) to better capture sub-carrier and timestamp information, effectively addressing the permutation-invariance problem. Furthermore, we propose a Mask Prediction Model (MPM) fine-tuning method to improve the model's adaptability for CSI prediction tasks. Experimental results demonstrate that CSI-BERT2 achieves state-of-the-art performance across all tasks.
Wireless sensing has recently found widespread applications in diverse environments, including homes, offices, and public spaces. By analyzing patterns in channel state information (CSI), it is possible to infer human actions for tasks such as person identification, gesture recognition, and fall detection. However, CSI is highly sensitive to environmental changes, where even minor alterations can significantly distort the CSI patterns. This sensitivity often leads to performance degradation or outright failure when applying wireless sensing models trained in one environment to another. To address this challenge, Domain Alignment (DAL) has been widely adopted for cross-domain classification tasks, as it focuses on aligning the global distributions of the source and target domains in feature space. Despite its popularity, DAL often neglects inter-category relationships, which can lead to misalignment between categories across domains, even when global alignment is achieved. To overcome these limitations, we propose K-Nearest Neighbors Maximum Mean Discrepancy (KNN-MMD), a novel few-shot method for cross-domain wireless sensing. Our approach begins by constructing a help set using KNN from the target domain, enabling local alignment between the source and target domains within each category using MMD. Additionally, we address a key instability issue commonly observed in cross-domain methods, where model performance fluctuates sharply between epochs. Further, most existing methods struggle to determine an optimal stopping point during training due to the absence of labeled data from the target domain. Our method resolves this by excluding the support set from the target domain during training and employing it as a validation set to determine the stopping criterion.
The human-robot interaction (HRI) is a growing area of research. In HRI, complex command (action) classification is still an open problem that usually prevents the real applicability of such a technique. The literature presents some works that use neural networks to detect these actions. However, occlusion is still a major issue in HRI, especially when using uncrewed aerial vehicles (UAVs), since, during the robot's movement, the human operator is often out of the robot's field of view. Furthermore, in multi-robot scenarios, distributed training is also an open problem. In this sense, this work proposes an action recognition and control approach based on Long Short-Term Memory (LSTM) Deep Neural Networks with two layers in association with three densely connected layers and Federated Learning (FL) embedded in multiple drones. The FL enabled our approach to be trained in a distributed fashion, i.e., access to data without the need for cloud or other repositories, which facilitates the multi-robot system's learning. Furthermore, our multi-robot approach results also prevented occlusion situations, with experiments with real robots achieving an accuracy greater than 96%.
As a globally celebrated sport, soccer has attracted widespread interest from fans all over the world. This paper aims to develop a comprehensive multi-modal framework for soccer video understanding. Specifically, we make the following contributions in this paper: (i) we introduce SoccerReplay-1988, the largest multi-modal soccer dataset to date, featuring videos and detailed annotations from 1,988 complete matches, with an automated annotation pipeline; (ii) we present the first visual-language foundation model in the soccer domain, MatchVision, which leverages spatiotemporal information across soccer videos and excels in various downstream tasks; (iii) we conduct extensive experiments and ablation studies on event classification, commentary generation, and multi-view foul recognition. MatchVision demonstrates state-of-the-art performance on all of them, substantially outperforming existing models, which highlights the superiority of our proposed data and model. We believe that this work will offer a standard paradigm for sports understanding research.