Maintaining a consistent identity across video frames while estimating pose is crucial for downstream tasks such as behavior analysis and action recognition, yet existing pipelines largely decouple tracking and pose estimation. Top-down methods rely on per-frame detections followed by identity association, leading to increased computational cost and error propagation, whereas bottom-up approaches often struggle to preserve identity reliably in multi-instance settings. To address these limitations, we introduce DeepSTEP (STEP), a Transformer-based framework for joint tracking and pose estimation across diverse animal species and humans. STEP unifies localization, identity maintenance, and keypoint estimation into a single model, eliminating the need for per-frame detections or post hoc matching. Our approach leverages discriminative model weight prediction to propagate target identity across frames while predicting anatomically meaningful keypoints. Discriminative model prediction typically requires predefined target states. In contrast, STEP assumes only an initial bounding box in the first frame, removing the need for explicit keypoint initialization. This is enabled by two components: Gaussian Map Soft Prediction (GMSP) and Offset Map Regression Adapter (OMRA), which allow direct keypoint prediction and simplify inference. Starting from the initial bounding box, STEP tracks the subject and estimates its pose in subsequent frames. We train and evaluate STEP on datasets spanning multiple species, demonstrating strong performance and greater efficiency than existing methods. By jointly modeling tracking and pose, STEP provides an effective and scalable solution.
Hand-object interaction detection remains an open challenge in real-time applications, where intuitive user experiences depend on fast and accurate detection of interactions with surrounding objects. We propose an efficient approach for detecting hand-objects interactions from streaming egocentric vision that operates in real time. Our approach consists of an action recognition module and an object detection module for identifying active objects upon confirmed interaction. Our Mamba model with EfficientNetV2 as backbone for action recognition achieves 38.52% p-AP on the ENIGMA-51 benchmark at 30fps, while our fine-tuned YOLOWorld reaches 85.13% AP for hand and object. We implement our models in a cascaded architecture where the action recognition and object detection modules operate sequentially. When the action recognition predicts a contact state, it activates the object detection module, which in turn performs inference on the relevant frame to detect and classify the active object.
Video understanding is fundamental to tasks such as action recognition, video reasoning, and robotic control. Early video understanding methods based on large vision-language models (LVLMs) typically adopt a single-pass reasoning paradigm without dynamic feedback, limiting the model's capacity to self-correct and adapt in complex scenarios. Recent efforts have attempted to address this limitation by incorporating reward models and reinforcement learning to enhance reasoning, or by employing tool-agent frameworks. However, these approaches face several challenges, including high annotation costs, reward signals that fail to capture real-time reasoning states, and low inference efficiency. To overcome these issues, we propose ReAgent-V, a novel agentic video understanding framework that integrates efficient frame selection with real-time reward generation during inference. These reward signals not only guide iterative answer refinement through a multi-perspective reflection mechanism-adjusting predictions from conservative, neutral, and aggressive viewpoints-but also enable automatic filtering of high-quality data for supervised fine-tuning (SFT), direct preference optimization (DPO), and group relative policy optimization (GRPO). ReAgent-V is lightweight, modular, and extensible, supporting flexible tool integration tailored to diverse tasks. Extensive experiments on 12 datasets across three core applications-video understanding, video reasoning enhancement, and vision-language-action model alignment-demonstrate significant gains in generalization and reasoning, with improvements of up to 6.9%, 2.1%, and 9.8%, respectively, highlighting the effectiveness and versatility of the proposed framework.
World models -- generative models that simulate environment dynamics conditioned on past observations and actions -- are gaining prominence in planning, simulation, and embodied AI. However, evaluating their rollouts remains a fundamental challenge, requiring fine-grained, temporally grounded assessment of action alignment and semantic consistency -- capabilities not captured by existing metrics. Vision-Language Models (VLMs) have shown promise as automatic evaluators of generative content due to their strong multimodal reasoning abilities. Yet, their use in fine-grained, temporally sensitive evaluation tasks remains limited and requires targeted adaptation. We introduce a evaluation protocol targeting two recognition tasks -- action recognition and character recognition -- each assessed across binary, multiple-choice, and open-ended formats. To support this, we present UNIVERSE (UNIfied Vision-language Evaluator for Rollouts in Simulated Environments), a method for adapting VLMs to rollout evaluation under data and compute constraints. We conduct a large-scale study comparing full, partial, and parameter-efficient finetuning across task formats, context lengths, sampling strategies, and data compositions. The resulting unified evaluator matches the performance of task-specific baselines using a single checkpoint. Human studies confirm strong alignment with human judgments, establishing UNIVERSE as a scalable, semantics-aware evaluator for world models.
Zero-shot skeleton-based action recognition aims to classify unseen skeleton-based human actions without prior exposure to such categories during training. This task is extremely challenging due to the difficulty in generalizing from known to unknown actions. Previous studies typically use two-stage training: pre-training skeleton encoders on seen action categories using cross-entropy loss and then aligning pre-extracted skeleton and text features, enabling knowledge transfer to unseen classes through skeleton-text alignment and language models' generalization. However, their efficacy is hindered by 1) insufficient discrimination for skeleton features, as the fixed skeleton encoder fails to capture necessary alignment information for effective skeleton-text alignment; 2) the neglect of alignment bias between skeleton and unseen text features during testing. To this end, we propose a prototype-guided feature alignment paradigm for zero-shot skeleton-based action recognition, termed PGFA. Specifically, we develop an end-to-end cross-modal contrastive training framework to improve skeleton-text alignment, ensuring sufficient discrimination for skeleton features. Additionally, we introduce a prototype-guided text feature alignment strategy to mitigate the adverse impact of the distribution discrepancy during testing. We provide a theoretical analysis to support our prototype-guided text feature alignment strategy and empirically evaluate our overall PGFA on three well-known datasets. Compared with the top competitor SMIE method, our PGFA achieves absolute accuracy improvements of 22.96%, 12.53%, and 18.54% on the NTU-60, NTU-120, and PKU-MMD datasets, respectively.
Effective human action recognition is widely used for cobots in Industry 4.0 to assist in assembly tasks. However, conventional skeleton-based methods often lose keypoint semantics, limiting their effectiveness in complex interactions. In this work, we introduce a novel approach to skeleton-based action recognition that enriches input representations by leveraging word embeddings to encode semantic information. Our method replaces one-hot encodings with semantic volumes, enabling the model to capture meaningful relationships between joints and objects. Through extensive experiments on multiple assembly datasets, we demonstrate that our approach significantly improves classification performance, and enhances generalization capabilities by simultaneously supporting different skeleton types and object classes. Our findings highlight the potential of incorporating semantic information to enhance skeleton-based action recognition in dynamic and diverse environments.
Modern perception models, particularly those designed for multisensory egocentric tasks, have achieved remarkable performance but often come with substantial computational costs. These high demands pose challenges for real-world deployment, especially in resource-constrained environments. In this paper, we introduce EgoAdapt, a framework that adaptively performs cross-modal distillation and policy learning to enable efficient inference across different egocentric perception tasks, including egocentric action recognition, active speaker localization, and behavior anticipation. Our proposed policy module is adaptable to task-specific action spaces, making it broadly applicable. Experimental results on three challenging egocentric datasets EPIC-Kitchens, EasyCom, and Aria Everyday Activities demonstrate that our method significantly enhances efficiency, reducing GMACs by up to 89.09%, parameters up to 82.02%, and energy up to 9.6x, while still on-par and in many cases outperforming, the performance of corresponding state-of-the-art models.
Owing to its rapid progress and broad application prospects, few-shot action recognition has attracted considerable interest. However, current methods are predominantly based on limited single-modal data, which does not fully exploit the potential of multimodal information. This paper presents a novel framework that actively identifies reliable modalities for each sample using task-specific contextual cues, thus significantly improving recognition performance. Our framework integrates an Active Sample Inference (ASI) module, which utilizes active inference to predict reliable modalities based on posterior distributions and subsequently organizes them accordingly. Unlike reinforcement learning, active inference replaces rewards with evidence-based preferences, making more stable predictions. Additionally, we introduce an active mutual distillation module that enhances the representation learning of less reliable modalities by transferring knowledge from more reliable ones. Adaptive multimodal inference is employed during the meta-test to assign higher weights to reliable modalities. Extensive experiments across multiple benchmarks demonstrate that our method significantly outperforms existing approaches.
We introduce CARMA, a system for situational grounding in human-robot group interactions. Effective collaboration in such group settings requires situational awareness based on a consistent representation of present persons and objects coupled with an episodic abstraction of events regarding actors and manipulated objects. This calls for a clear and consistent assignment of instances, ensuring that robots correctly recognize and track actors, objects, and their interactions over time. To achieve this, CARMA uniquely identifies physical instances of such entities in the real world and organizes them into grounded triplets of actors, objects, and actions. To validate our approach, we conducted three experiments, where multiple humans and a robot interact: collaborative pouring, handovers, and sorting. These scenarios allow the assessment of the system's capabilities as to role distinction, multi-actor awareness, and consistent instance identification. Our experiments demonstrate that the system can reliably generate accurate actor-action-object triplets, providing a structured and robust foundation for applications requiring spatiotemporal reasoning and situated decision-making in collaborative settings.
Understanding human actions in videos requires more than raw pixel analysis; it relies on high-level semantic reasoning and effective integration of multimodal features. We propose a deep translational action recognition framework that enhances recognition accuracy by jointly predicting action concepts and auxiliary features from RGB video frames. At test time, hallucination streams infer missing cues, enriching feature representations without increasing computational overhead. To focus on action-relevant regions beyond raw pixels, we introduce two novel domain-specific descriptors. Object Detection Features (ODF) aggregate outputs from multiple object detectors to capture contextual cues, while Saliency Detection Features (SDF) highlight spatial and intensity patterns crucial for action recognition. Our framework seamlessly integrates these descriptors with auxiliary modalities such as optical flow, Improved Dense Trajectories, skeleton data, and audio cues. It remains compatible with state-of-the-art architectures, including I3D, AssembleNet, Video Transformer Network, FASTER, and recent models like VideoMAE V2 and InternVideo2. To handle uncertainty in auxiliary features, we incorporate aleatoric uncertainty modeling in the hallucination step and introduce a robust loss function to mitigate feature noise. Our multimodal self-supervised action recognition framework achieves state-of-the-art performance on multiple benchmarks, including Kinetics-400, Kinetics-600, and Something-Something V2, demonstrating its effectiveness in capturing fine-grained action dynamics.
Understanding behavior requires datasets that capture humans while carrying out complex tasks. The kitchen is an excellent environment for assessing human motor and cognitive function, as many complex actions are naturally exhibited in kitchens from chopping to cleaning. Here, we introduce the EPFL-Smart-Kitchen-30 dataset, collected in a noninvasive motion capture platform inside a kitchen environment. Nine static RGB-D cameras, inertial measurement units (IMUs) and one head-mounted HoloLens~2 headset were used to capture 3D hand, body, and eye movements. The EPFL-Smart-Kitchen-30 dataset is a multi-view action dataset with synchronized exocentric, egocentric, depth, IMUs, eye gaze, body and hand kinematics spanning 29.7 hours of 16 subjects cooking four different recipes. Action sequences were densely annotated with 33.78 action segments per minute. Leveraging this multi-modal dataset, we propose four benchmarks to advance behavior understanding and modeling through 1) a vision-language benchmark, 2) a semantic text-to-motion generation benchmark, 3) a multi-modal action recognition benchmark, 4) a pose-based action segmentation benchmark. We expect the EPFL-Smart-Kitchen-30 dataset to pave the way for better methods as well as insights to understand the nature of ecologically-valid human behavior. Code and data are available at https://github.com/amathislab/EPFL-Smart-Kitchen
In many complex scenarios, robotic manipulation relies on generative models to estimate the distribution of multiple successful actions. As the diffusion model has better training robustness than other generative models, it performs well in imitation learning through successful robot demonstrations. However, the diffusion-based policy methods typically require significant time to iteratively denoise robot actions, which hinders real-time responses in robotic manipulation. Moreover, existing diffusion policies model a time-varying action denoising process, whose temporal complexity increases the difficulty of model training and leads to suboptimal action accuracy. To generate robot actions efficiently and accurately, we present the Time-Unified Diffusion Policy (TUDP), which utilizes action recognition capabilities to build a time-unified denoising process. On the one hand, we build a time-unified velocity field in action space with additional action discrimination information. By unifying all timesteps of action denoising, our velocity field reduces the difficulty of policy learning and speeds up action generation. On the other hand, we propose an action-wise training method, which introduces an action discrimination branch to supply additional action discrimination information. Through action-wise training, the TUDP implicitly learns the ability to discern successful actions to better denoising accuracy. Our method achieves state-of-the-art performance on RLBench with the highest success rate of 82.6% on a multi-view setup and 83.8% on a single-view setup. In particular, when using fewer denoising iterations, TUDP achieves a more significant improvement in success rate. Additionally, TUDP can produce accurate actions for a wide range of real-world tasks.
Action segmentation is a core challenge in high-level video understanding, aiming to partition untrimmed videos into segments and assign each a label from a predefined action set. Existing methods primarily address single-person activities with fixed action sequences, overlooking multi-person scenarios. In this work, we pioneer textual reference-guided human action segmentation in multi-person settings, where a textual description specifies the target person for segmentation. We introduce the first dataset for Referring Human Action Segmentation, i.e., RHAS133, built from 133 movies and annotated with 137 fine-grained actions with 33h video data, together with textual descriptions for this new task. Benchmarking existing action recognition methods on RHAS133 using VLM-based feature extractors reveals limited performance and poor aggregation of visual cues for the target person. To address this, we propose a holistic-partial aware Fourier-conditioned diffusion framework, i.e., HopaDIFF, leveraging a novel cross-input gate attentional xLSTM to enhance holistic-partial long-range reasoning and a novel Fourier condition to introduce more fine-grained control to improve the action segmentation generation. HopaDIFF achieves state-of-the-art results on RHAS133 in diverse evaluation settings. The code is available at https://github.com/KPeng9510/HopaDIFF.git.
Recently, multimodal tasks have strongly advanced the field of action recognition with their rich multimodal information. However, due to the scarcity of tri-modal data, research on tri-modal action recognition tasks faces many challenges. To this end, we have proposed a comprehensive multimodal action recognition solution that effectively utilizes multimodal information. First, the existing data are transformed and expanded by optimizing data enhancement techniques to enlarge the training scale. At the same time, more RGB datasets are used to pre-train the backbone network, which is better adapted to the new task by means of transfer learning. Secondly, multimodal spatial features are extracted with the help of 2D CNNs and combined with the Temporal Shift Module (TSM) to achieve multimodal spatial-temporal feature extraction comparable to 3D CNNs and improve the computational efficiency. In addition, common prediction enhancement methods, such as Stochastic Weight Averaging (SWA), Ensemble and Test-Time augmentation (TTA), are used to integrate the knowledge of models from different training periods of the same architecture and different architectures, so as to predict the actions from different perspectives and fully exploit the target information. Ultimately, we achieved the Top-1 accuracy of 99% and the Top-5 accuracy of 100% on the competition leaderboard, demonstrating the superiority of our solution.
Few-Shot Action Recognition (FSAR) aims to train a model with only a few labeled video instances. A key challenge in FSAR is handling divergent narrative trajectories for precise video matching. While the frame- and tuple-level alignment approaches have been promising, their methods heavily rely on pre-defined and length-dependent alignment units (e.g., frames or tuples), which limits flexibility for actions of varying lengths and speeds. In this work, we introduce a novel TEmporal Alignment-free Matching (TEAM) approach, which eliminates the need for temporal units in action representation and brute-force alignment during matching. Specifically, TEAM represents each video with a fixed set of pattern tokens that capture globally discriminative clues within the video instance regardless of action length or speed, ensuring its flexibility. Furthermore, TEAM is inherently efficient, using token-wise comparisons to measure similarity between videos, unlike existing methods that rely on pairwise comparisons for temporal alignment. Additionally, we propose an adaptation process that identifies and removes common information across classes, establishing clear boundaries even between novel categories. Extensive experiments demonstrate the effectiveness of TEAM. Codes are available at github.com/leesb7426/TEAM.
This review provides an in-depth exploration of the field of animal action recognition, focusing on coarse-grained (CG) and fine-grained (FG) techniques. The primary aim is to examine the current state of research in animal behaviour recognition and to elucidate the unique challenges associated with recognising subtle animal actions in outdoor environments. These challenges differ significantly from those encountered in human action recognition due to factors such as non-rigid body structures, frequent occlusions, and the lack of large-scale, annotated datasets. The review begins by discussing the evolution of human action recognition, a more established field, highlighting how it progressed from broad, coarse actions in controlled settings to the demand for fine-grained recognition in dynamic environments. This shift is particularly relevant for animal action recognition, where behavioural variability and environmental complexity present unique challenges that human-centric models cannot fully address. The review then underscores the critical differences between human and animal action recognition, with an emphasis on high intra-species variability, unstructured datasets, and the natural complexity of animal habitats. Techniques like spatio-temporal deep learning frameworks (e.g., SlowFast) are evaluated for their effectiveness in animal behaviour analysis, along with the limitations of existing datasets. By assessing the strengths and weaknesses of current methodologies and introducing a recently-published dataset, the review outlines future directions for advancing fine-grained action recognition, aiming to improve accuracy and generalisability in behaviour analysis across species.
With the inherent advantages of skeleton representation, 3D skeleton-based action recognition has become a prominent topic in the field of computer vision. However, previous reviews have predominantly adopted a model-oriented perspective, often neglecting the fundamental steps involved in skeleton-based action recognition. This oversight tends to ignore key components of skeleton-based action recognition beyond model design and has hindered deeper, more intrinsic understanding of the task. To bridge this gap, our review aims to address these limitations by presenting a comprehensive, task-oriented framework for understanding skeleton-based action recognition. We begin by decomposing the task into a series of sub-tasks, placing particular emphasis on preprocessing steps such as modality derivation and data augmentation. The subsequent discussion delves into critical sub-tasks, including feature extraction and spatio-temporal modeling techniques. Beyond foundational action recognition networks, recently advanced frameworks such as hybrid architectures, Mamba models, large language models (LLMs), and generative models have also been highlighted. Finally, a comprehensive overview of public 3D skeleton datasets is presented, accompanied by an analysis of state-of-the-art algorithms evaluated on these benchmarks. By integrating task-oriented discussions, comprehensive examinations of sub-tasks, and an emphasis on the latest advancements, our review provides a fundamental and accessible structured roadmap for understanding and advancing the field of 3D skeleton-based action recognition.
Operating rooms (ORs) demand precise coordination among surgeons, nurses, and equipment in a fast-paced, occlusion-heavy environment, necessitating advanced perception models to enhance safety and efficiency. Existing datasets either provide partial egocentric views or sparse exocentric multi-view context, but do not explore the comprehensive combination of both. We introduce EgoExOR, the first OR dataset and accompanying benchmark to fuse first-person and third-person perspectives. Spanning 94 minutes (84,553 frames at 15 FPS) of two emulated spine procedures, Ultrasound-Guided Needle Insertion and Minimally Invasive Spine Surgery, EgoExOR integrates egocentric data (RGB, gaze, hand tracking, audio) from wearable glasses, exocentric RGB and depth from RGB-D cameras, and ultrasound imagery. Its detailed scene graph annotations, covering 36 entities and 22 relations (568,235 triplets), enable robust modeling of clinical interactions, supporting tasks like action recognition and human-centric perception. We evaluate the surgical scene graph generation performance of two adapted state-of-the-art models and offer a new baseline that explicitly leverages EgoExOR's multimodal and multi-perspective signals. This new dataset and benchmark set a new foundation for OR perception, offering a rich, multimodal resource for next-generation clinical perception.
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.
Action recognition models using deep learning are vulnerable to adversarial examples, which are transferable across other models trained on the same data modality. Existing transferable attack methods face two major challenges: 1) they heavily rely on the assumption that the decision boundaries of the surrogate (a.k.a., source) model and the target model are similar, which limits the adversarial transferability; and 2) their decision boundary difference makes the attack direction uncertain, which may result in the gradient oscillation, weakening the adversarial attack. This motivates us to propose a Background Mixup-induced Temporal Consistency (BMTC) attack method for action recognition. From the input transformation perspective, we design a model-agnostic background adversarial mixup module to reduce the surrogate-target model dependency. In particular, we randomly sample one video from each category and make its background frame, while selecting the background frame with the top attack ability for mixup with the clean frame by reinforcement learning. Moreover, to ensure an explicit attack direction, we leverage the background category as guidance for updating the gradient of adversarial example, and design a temporal gradient consistency loss, which strengthens the stability of the attack direction on subsequent frames. Empirical studies on two video datasets, i.e., UCF101 and Kinetics-400, and one image dataset, i.e., ImageNet, demonstrate that our method significantly boosts the transferability of adversarial examples across several action/image recognition models. Our code is available at https://github.com/mlvccn/BMTC_TransferAttackVid.