In this paper, we present a one-stage framework TriDet for temporal action detection. Existing methods often suffer from imprecise boundary predictions due to the ambiguous action boundaries in videos. To alleviate this problem, we propose a novel Trident-head to model the action boundary via an estimated relative probability distribution around the boundary. In the feature pyramid of TriDet, we propose an efficient Scalable-Granularity Perception (SGP) layer to mitigate the rank loss problem of self-attention that takes place in the video features and aggregate information across different temporal granularities. Benefiting from the Trident-head and the SGP-based feature pyramid, TriDet achieves state-of-the-art performance on three challenging benchmarks: THUMOS14, HACS and EPIC-KITCHEN 100, with lower computational costs, compared to previous methods. For example, TriDet hits an average mAP of $69.3\%$ on THUMOS14, outperforming the previous best by $2.5\%$, but with only $74.6\%$ of its latency. The code is released to https://github.com/sssste/TriDet.
Human action recognition aims at classifying the category of human action from a segment of a video. Recently, people have dived into designing GCN-based models to extract features from skeletons for performing this task, because skeleton representations are much more efficient and robust than other modalities such as RGB frames. However, when employing the skeleton data, some important clues like related items are also discarded. It results in some ambiguous actions that are hard to be distinguished and tend to be misclassified. To alleviate this problem, we propose an auxiliary feature refinement head (FR Head), which consists of spatial-temporal decoupling and contrastive feature refinement, to obtain discriminative representations of skeletons. Ambiguous samples are dynamically discovered and calibrated in the feature space. Furthermore, FR Head could be imposed on different stages of GCNs to build a multi-level refinement for stronger supervision. Extensive experiments are conducted on NTU RGB+D, NTU RGB+D 120, and NW-UCLA datasets. Our proposed models obtain competitive results from state-of-the-art methods and can help to discriminate those ambiguous samples. Codes are available at https://github.com/zhysora/FR-Head.
We present a novel approach for action recognition in UAV videos. Our formulation is designed to handle occlusion and viewpoint changes caused by the movement of a UAV. We use the concept of mutual information to compute and align the regions corresponding to human action or motion in the temporal domain. This enables our recognition model to learn from the key features associated with the motion. We also propose a novel frame sampling method that uses joint mutual information to acquire the most informative frame sequence in UAV videos. We have integrated our approach with X3D and evaluated the performance on multiple datasets. In practice, we achieve 18.9% improvement in Top-1 accuracy over current state-of-the-art methods on UAV-Human(Li et al., 2021), 7.3% improvement on Drone-Action(Perera et al., 2019), and 7.16% improvement on NEC Drones(Choi et al., 2020).
Temporal action localization in videos presents significant challenges in the field of computer vision. While the boundary-sensitive method has been widely adopted, its limitations include incomplete use of intermediate and global information, as well as an inefficient proposal feature generator. To address these challenges, we propose a novel framework, Sparse Multilevel Boundary Generator (SMBG), which enhances the boundary-sensitive method with boundary classification and action completeness regression. SMBG features a multi-level boundary module that enables faster processing by gathering boundary information at different lengths. Additionally, we introduce a sparse extraction confidence head that distinguishes information inside and outside the action, further optimizing the proposal feature generator. To improve the synergy between multiple branches and balance positive and negative samples, we propose a global guidance loss. Our method is evaluated on two popular benchmarks, ActivityNet-1.3 and THUMOS14, and is shown to achieve state-of-the-art performance, with a better inference speed (2.47xBSN++, 2.12xDBG). These results demonstrate that SMBG provides a more efficient and simple solution for generating temporal action proposals. Our proposed framework has the potential to advance the field of computer vision and enhance the accuracy and speed of temporal action localization in video analysis.The code and models are made available at \url{https://github.com/zhouyang-001/SMBG-for-temporal-action-proposal}.
We propose a novel approach for aerial video action recognition. Our method is designed for videos captured using UAVs and can run on edge or mobile devices. We present a learning-based approach that uses customized auto zoom to automatically identify the human target and scale it appropriately. This makes it easier to extract the key features and reduces the computational overhead. We also present an efficient temporal reasoning algorithm to capture the action information along the spatial and temporal domains within a controllable computational cost. Our approach has been implemented and evaluated both on the desktop with high-end GPUs and on the low power Robotics RB5 Platform for robots and drones. In practice, we achieve 6.1-7.4% improvement over SOTA in Top-1 accuracy on the RoCoG-v2 dataset, 8.3-10.4% improvement on the UAV-Human dataset and 3.2% improvement on the Drone Action dataset.
The performance of video action recognition has been significantly boosted by using motion representations within a two-stream Convolutional Neural Network (CNN) architecture. However, there are a few challenging problems in action recognition in real scenarios, e.g., the variations in viewpoints and poses, and the changes in backgrounds. The domain discrepancy between the training data and the test data causes the performance drop. To improve the model robustness, we propose a novel method to determine the task-irrelevant content in inputs which increases the domain discrepancy. The method is based on a human parsing model (HP model) which jointly conducts dense correspondence labelling and semantic part segmentation. The predictions from the HP model also function as re-rendering the human regions in each video using the same set of textures to make humans appearances in all classes be the same. A revised dataset is generated for training and testing and makes the action recognition model exhibit invariance to the irrelevant content in the inputs. Moreover, the predictions from the HP model are used to enrich the inputs to the AR model during both training and testing. Experimental results show that our proposed model is superior to existing models for action recognition on the HMDB-51 dataset and the Penn Action dataset.
Recognizing human actions in video sequences, known as Human Action Recognition (HAR), is a challenging task in pattern recognition. While Convolutional Neural Networks (ConvNets) have shown remarkable success in image recognition, they are not always directly applicable to HAR, as temporal features are critical for accurate classification. In this paper, we propose a novel dynamic PSO-ConvNet model for learning actions in videos, building on our recent work in image recognition. Our approach leverages a framework where the weight vector of each neural network represents the position of a particle in phase space, and particles share their current weight vectors and gradient estimates of the Loss function. To extend our approach to video, we integrate ConvNets with state-of-the-art temporal methods such as Transformer and Recurrent Neural Networks. Our experimental results on the UCF-101 dataset demonstrate substantial improvements of up to 9% in accuracy, which confirms the effectiveness of our proposed method. In addition, we conducted experiments on larger and more variety of datasets including Kinetics-400 and HMDB-51 and obtained preference for Collaborative Learning in comparison with Non-Collaborative Learning (Individual Learning). Overall, our dynamic PSO-ConvNet model provides a promising direction for improving HAR by better capturing the spatio-temporal dynamics of human actions in videos. The code is available at https://github.com/leonlha/Video-Action-Recognition-Collaborative-Learning-with-Dynamics-via-PSO-ConvNet-Transformer.
In this work, the novel task of detecting and classifying table tennis strokes solely using the ball trajectory has been explored. A single camera setup positioned in the umpire's view has been employed to procure a dataset consisting of six stroke classes executed by four professional table tennis players. Ball tracking using YOLOv4, a traditional object detection model, and TrackNetv2, a temporal heatmap based model, have been implemented on our dataset and their performances have been benchmarked. A mathematical approach developed to extract temporal boundaries of strokes using the ball trajectory data yielded a total of 2023 valid strokes in our dataset, while also detecting services and missed strokes successfully. The temporal convolutional network developed performed stroke recognition on completely unseen data with an accuracy of 87.155%. Several machine learning and deep learning based model architectures have been trained for stroke recognition using ball trajectory input and benchmarked based on their performances. While stroke recognition in the field of table tennis has been extensively explored based on human action recognition using video data focused on the player's actions, the use of ball trajectory data for the same is an unexplored characteristic of the sport. Hence, the motivation behind the work is to demonstrate that meaningful inferences such as stroke detection and recognition can be drawn using minimal input information.
Recent vision transformer based video models mostly follow the ``image pre-training then finetuning" paradigm and have achieved great success on multiple video benchmarks. However, full finetuning such a video model could be computationally expensive and unnecessary, given the pre-trained image transformer models have demonstrated exceptional transferability. In this work, we propose a novel method to Adapt pre-trained Image Models (AIM) for efficient video understanding. By freezing the pre-trained image model and adding a few lightweight Adapters, we introduce spatial adaptation, temporal adaptation and joint adaptation to gradually equip an image model with spatiotemporal reasoning capability. We show that our proposed AIM can achieve competitive or even better performance than prior arts with substantially fewer tunable parameters on four video action recognition benchmarks. Thanks to its simplicity, our method is also generally applicable to different image pre-trained models, which has the potential to leverage more powerful image foundation models in the future. The project webpage is \url{https://adapt-image-models.github.io/}.
Recently, there has been a remarkable increase in the interest towards skeleton-based action recognition within the research community, owing to its various advantageous features, including computational efficiency, representative features, and illumination invariance. Despite this, researchers continue to explore and investigate the most optimal way to represent human actions through skeleton representation and the extracted features. As a result, the growth and availability of human action recognition datasets have risen substantially. In addition, deep learning-based algorithms have gained widespread popularity due to the remarkable advancements in various computer vision tasks. Most state-of-the-art contributions in skeleton-based action recognition incorporate a Graph Neural Network (GCN) architecture for representing the human body and extracting features. Our research demonstrates that Convolutional Neural Networks (CNNs) can attain comparable results to GCN, provided that the proper training techniques, augmentations, and optimizers are applied. Our approach has been rigorously validated, and we have achieved a score of 95% on the NTU-60 dataset
Due to the compact and rich high-level representations offered, skeleton-based human action recognition has recently become a highly active research topic. Previous studies have demonstrated that investigating joint relationships in spatial and temporal dimensions provides effective information critical to action recognition. However, effectively encoding global dependencies of joints during spatio-temporal feature extraction is still challenging. In this paper, we introduce Action Capsule which identifies action-related key joints by considering the latent correlation of joints in a skeleton sequence. We show that, during inference, our end-to-end network pays attention to a set of joints specific to each action, whose encoded spatio-temporal features are aggregated to recognize the action. Additionally, the use of multiple stages of action capsules enhances the ability of the network to classify similar actions. Consequently, our network outperforms the state-of-the-art approaches on the N-UCLA dataset and obtains competitive results on the NTURGBD dataset. This is while our approach has significantly lower computational requirements based on GFLOPs measurements.
Video-Language Pre-training models have recently significantly improved various multi-modal downstream tasks. Previous dominant works mainly adopt contrastive learning to achieve global feature alignment across modalities. However, the local associations between videos and texts are not modeled, restricting the pre-training models' generality, especially for tasks requiring the temporal video boundary for certain query texts. This work introduces a novel text-video localization pre-text task to enable fine-grained temporal and semantic alignment such that the trained model can accurately perceive temporal boundaries in videos given the text description. Specifically, text-video localization consists of moment retrieval, which predicts start and end boundaries in videos given the text description, and text localization which matches the subset of texts with the video features. To produce temporal boundaries, frame features in several videos are manually merged into a long video sequence that interacts with a text sequence. With the localization task, our method connects the fine-grained frame representations with the word representations and implicitly distinguishes representations of different instances in the single modality. Notably, comprehensive experimental results show that our method significantly improves the state-of-the-art performance on various benchmarks, covering text-to-video retrieval, video question answering, video captioning, temporal action localization and temporal moment retrieval. The code will be released soon.
Human Action Recognition (HAR), one of the most important tasks in computer vision, has developed rapidly in the past decade and has a wide range of applications in health monitoring, intelligent surveillance, virtual reality, human computer interaction and so on. Human actions can be represented by a wide variety of modalities, such as RGB-D cameras, audio, inertial sensors,etc. Consequently, in addition to the mainstream single modality based HAR approaches, more and more research is devoted to the multimodal domain due to the complementary properties between multimodal data. In this paper, we present a survey of HAR methods in recent years according to the different input modalities. Meanwhile, considering that most of the recent surveys on HAR focus on the third perspective, while this survey aims to provide a more comprehensive introduction to HAR novices and researchers, we therefore also investigate the actions recognition methods from the first perspective in recent years. Finally, we give a brief introduction about the benchmark HAR datasets and show the performance comparison of different methods on these datasets.
Contrastive Masked Autoencoder (CMAE), as a new self-supervised framework, has shown its potential of learning expressive feature representations in visual image recognition. This work shows that CMAE also trivially generalizes well on video action recognition without modifying the architecture and the loss criterion. By directly replacing the original pixel shift with the temporal shift, our CMAE for visual action recognition, CMAE-V for short, can generate stronger feature representations than its counterpart based on pure masked autoencoders. Notably, CMAE-V, with a hybrid architecture, can achieve 82.2% and 71.6% top-1 accuracy on the Kinetics-400 and Something-something V2 datasets, respectively. We hope this report could provide some informative inspiration for future works.
Classifying the behavior of humans or animals from videos is important in biomedical fields for understanding brain function and response to stimuli. Action recognition, classifying activities performed by one or more subjects in a trimmed video, forms the basis of many of these techniques. Deep learning models for human action recognition have progressed significantly over the last decade. Recently, there is an increased interest in research that incorporates deep learning-based action recognition for animal behavior classification. However, human action recognition methods are more developed. This survey presents an overview of human action recognition and pose estimation methods that are based on convolutional neural network (CNN) architectures and have been adapted for animal behavior classification in neuroscience. Pose estimation, estimating joint positions from an image frame, is included because it is often applied before classifying animal behavior. First, we provide foundational information on algorithms that learn spatiotemporal features through 2D, two-stream, and 3D CNNs. We explore motivating factors that determine optimizers, loss functions and training procedures, and compare their performance on benchmark datasets. Next, we review animal behavior frameworks that use or build upon these methods, organized by the level of supervision they require. Our discussion is uniquely focused on the technical evolution of the underlying CNN models and their architectural adaptations (which we illustrate), rather than their usability in a neuroscience lab. We conclude by discussing open research problems, and possible research directions. Our survey is designed to be a resource for researchers developing fully unsupervised animal behavior classification systems of which there are only a few examples in the literature.
We present Ego-Only, the first approach that enables state-of-the-art action detection on egocentric (first-person) videos without any form of exocentric (third-person) transferring. Despite the content and appearance gap separating the two domains, large-scale exocentric transferring has been the default choice for egocentric action detection. This is because prior works found that egocentric models are difficult to train from scratch and that transferring from exocentric representations leads to improved accuracy. However, in this paper, we revisit this common belief. Motivated by the large gap separating the two domains, we propose a strategy that enables effective training of egocentric models without exocentric transferring. Our Ego-Only approach is simple. It trains the video representation with a masked autoencoder finetuned for temporal segmentation. The learned features are then fed to an off-the-shelf temporal action localization method to detect actions. We find that this renders exocentric transferring unnecessary by showing remarkably strong results achieved by this simple Ego-Only approach on three established egocentric video datasets: Ego4D, EPIC-Kitchens-100, and Charades-Ego. On both action detection and action recognition, Ego-Only outperforms previous best exocentric transferring methods that use orders of magnitude more labels. Ego-Only sets new state-of-the-art results on these datasets and benchmarks without exocentric data.
Deep neural networks (DNNs) are vulnerable to a class of attacks called "backdoor attacks", which create an association between a backdoor trigger and a target label the attacker is interested in exploiting. A backdoored DNN performs well on clean test images, yet persistently predicts an attacker-defined label for any sample in the presence of the backdoor trigger. Although backdoor attacks have been extensively studied in the image domain, there are very few works that explore such attacks in the video domain, and they tend to conclude that image backdoor attacks are less effective in the video domain. In this work, we revisit the traditional backdoor threat model and incorporate additional video-related aspects to that model. We show that poisoned-label image backdoor attacks could be extended temporally in two ways, statically and dynamically, leading to highly effective attacks in the video domain. In addition, we explore natural video backdoors to highlight the seriousness of this vulnerability in the video domain. And, for the first time, we study multi-modal (audiovisual) backdoor attacks against video action recognition models, where we show that attacking a single modality is enough for achieving a high attack success rate.
To interpret deep neural networks, one main approach is to dissect the visual input and find the prototypical parts responsible for the classification. However, existing methods often ignore the hierarchical relationship between these prototypes, and thus can not explain semantic concepts at both higher level (e.g., water sports) and lower level (e.g., swimming). In this paper inspired by human cognition system, we leverage hierarchal information to deal with uncertainty: When we observe water and human activity, but no definitive action it can be recognized as the water sports parent class. Only after observing a person swimming can we definitively refine it to the swimming action. To this end, we propose HIerarchical Prototype Explainer (HIPE) to build hierarchical relations between prototypes and classes. HIPE enables a reasoning process for video action classification by dissecting the input video frames on multiple levels of the class hierarchy, our method is also applicable to other video tasks. The faithfulness of our method is verified by reducing accuracy-explainability trade off on ActivityNet and UCF-101 while providing multi-level explanations.
Previous work has shown the potential of deep learning to predict renal obstruction using kidney ultrasound images. However, these image-based classifiers have been trained with the goal of single-visit inference in mind. We compare methods from video action recognition (i.e. convolutional pooling, LSTM, TSM) to adapt single-visit convolutional models to handle multiple visit inference. We demonstrate that incorporating images from a patient's past hospital visits provides only a small benefit for the prediction of obstructive hydronephrosis. Therefore, inclusion of prior ultrasounds is beneficial, but prediction based on the latest ultrasound is sufficient for patient risk stratification.
In recent years multi-label, multi-class video action recognition has gained significant popularity. While reasoning over temporally connected atomic actions is mundane for intelligent species, standard artificial neural networks (ANN) still struggle to classify them. In the real world, atomic actions often temporally connect to form more complex composite actions. The challenge lies in recognising composite action of varying durations while other distinct composite or atomic actions occur in the background. Drawing upon the success of relational networks, we propose methods that learn to reason over the semantic concept of objects and actions. We empirically show how ANNs benefit from pretraining, relational inductive biases and unordered set-based latent representations. In this paper we propose deep set conditioned I3D (SCI3D), a two stream relational network that employs latent representation of state and visual representation for reasoning over events and actions. They learn to reason about temporally connected actions in order to identify all of them in the video. The proposed method achieves an improvement of around 1.49% mAP in atomic action recognition and 17.57% mAP in composite action recognition, over a I3D-NL baseline, on the CATER dataset.