Most modern approaches in temporal action localization divide this problem into two parts: (i) short-term feature extraction and (ii) long-range temporal boundary localization. Due to the high GPU memory cost caused by processing long untrimmed videos, many methods sacrifice the representational power of the short-term feature extractor by either freezing the backbone or using a small spatial video resolution. This issue becomes even worse with the recent video transformer models, many of which have quadratic memory complexity. To address these issues, we propose TALLFormer, a memory-efficient and end-to-end trainable Temporal Action Localization Transformer with Long-term memory. Our long-term memory mechanism eliminates the need for processing hundreds of redundant video frames during each training iteration, thus, significantly reducing the GPU memory consumption and training time. These efficiency savings allow us (i) to use a powerful video transformer feature extractor without freezing the backbone or reducing the spatial video resolution, while (ii) also maintaining long-range temporal boundary localization capability. With only RGB frames as input and no external action recognition classifier, TALLFormer outperforms previous state-of-the-arts by a large margin, achieving an average mAP of 59.1% on THUMOS14 and 35.6% on ActivityNet-1.3. The code is public available: https://github.com/klauscc/TALLFormer.
We target at the task of weakly-supervised action localization (WSAL), where only video-level action labels are available during model training. Despite the recent progress, existing methods mainly embrace a localization-by-classification paradigm and overlook the fruitful fine-grained temporal distinctions between video sequences, thus suffering from severe ambiguity in classification learning and classification-to-localization adaption. This paper argues that learning by contextually comparing sequence-to-sequence distinctions offers an essential inductive bias in WSAL and helps identify coherent action instances. Specifically, under a differentiable dynamic programming formulation, two complementary contrastive objectives are designed, including Fine-grained Sequence Distance (FSD) contrasting and Longest Common Subsequence (LCS) contrasting, where the first one considers the relations of various action/background proposals by using match, insert, and delete operators and the second one mines the longest common subsequences between two videos. Both contrasting modules can enhance each other and jointly enjoy the merits of discriminative action-background separation and alleviated task gap between classification and localization. Extensive experiments show that our method achieves state-of-the-art performance on two popular benchmarks. Our code is available at https://github.com/MengyuanChen21/CVPR2022-FTCL.
Weakly-supervised temporal action localization aims to recognize and localize action segments in untrimmed videos given only video-level action labels for training. Without the boundary information of action segments, existing methods mostly rely on multiple instance learning (MIL), where the predictions of unlabeled instances (i.e., video snippets) are supervised by classifying labeled bags (i.e., untrimmed videos). However, this formulation typically treats snippets in a video as independent instances, ignoring the underlying temporal structures within and across action segments. To address this problem, we propose \system, a novel WTAL framework that enables explicit, action-aware segment modeling beyond standard MIL-based methods. Our framework entails three segment-centric components: (i) dynamic segment sampling for compensating the contribution of short actions; (ii) intra- and inter-segment attention for modeling action dynamics and capturing temporal dependencies; (iii) pseudo instance-level supervision for improving action boundary prediction. Furthermore, a multi-step refinement strategy is proposed to progressively improve action proposals along the model training process. Extensive experiments on THUMOS-14 and ActivityNet-v1.3 demonstrate the effectiveness of our approach, establishing new state of the art on both datasets. The code and models are publicly available at~\url{https://github.com/boheumd/ASM-Loc}.
Unsupervised video representation learning has made remarkable achievements in recent years. However, most existing methods are designed and optimized for video classification. These pre-trained models can be sub-optimal for temporal localization tasks due to the inherent discrepancy between video-level classification and clip-level localization. To bridge this gap, we make the first attempt to propose a self-supervised pretext task, coined as Pseudo Action Localization (PAL) to Unsupervisedly Pre-train feature encoders for Temporal Action Localization tasks (UP-TAL). Specifically, we first randomly select temporal regions, each of which contains multiple clips, from one video as pseudo actions and then paste them onto different temporal positions of the other two videos. The pretext task is to align the features of pasted pseudo action regions from two synthetic videos and maximize the agreement between them. Compared to the existing unsupervised video representation learning approaches, our PAL adapts better to downstream TAL tasks by introducing a temporal equivariant contrastive learning paradigm in a temporally dense and scale-aware manner. Extensive experiments show that PAL can utilize large-scale unlabeled video data to significantly boost the performance of existing TAL methods. Our codes and models will be made publicly available at https://github.com/zhang-can/UP-TAL.
Human action recognition has recently become one of the popular research topics in the computer vision community. Various 3D-CNN based methods have been presented to tackle both the spatial and temporal dimensions in the task of video action recognition with competitive results. However, these methods have suffered some fundamental limitations such as lack of robustness and generalization, e.g., how does the temporal ordering of video frames affect the recognition results? This work presents a novel end-to-end Transformer-based Directed Attention (DirecFormer) framework for robust action recognition. The method takes a simple but novel perspective of Transformer-based approach to understand the right order of sequence actions. Therefore, the contributions of this work are three-fold. Firstly, we introduce the problem of ordered temporal learning issues to the action recognition problem. Secondly, a new Directed Attention mechanism is introduced to understand and provide attentions to human actions in the right order. Thirdly, we introduce the conditional dependency in action sequence modeling that includes orders and classes. The proposed approach consistently achieves the state-of-the-art (SOTA) results compared with the recent action recognition methods, on three standard large-scale benchmarks, i.e. Jester, Kinetics-400 and Something-Something-V2.
Graph-based reasoning over skeleton data has emerged as a promising approach for human action recognition. However, the application of prior graph-based methods, which predominantly employ whole temporal sequences as their input, to the setting of online inference entails considerable computational redundancy. In this paper, we tackle this issue by reformulating the Spatio-Temporal Graph Convolutional Neural Network as a Continual Inference Network, which can perform step-by-step predictions in time without repeat frame processing. To evaluate our method, we create a continual version of ST-GCN, CoST-GCN, alongside two derived methods with different self-attention mechanisms, CoAGCN and CoS-TR. We investigate weight transfer strategies and architectural modifications for inference acceleration, and perform experiments on the NTU RGB+D 60, NTU RGB+D 120, and Kinetics Skeleton 400 datasets. Retaining similar predictive accuracy, we observe up to 109x reduction in time complexity, on-hardware accelerations of 26x, and reductions in maximum allocated memory of 52% during online inference.
Convolutional Neural Networks are the de facto models for image recognition. However 3D CNNs, the straight forward extension of 2D CNNs for video recognition, have not achieved the same success on standard action recognition benchmarks. One of the main reasons for this reduced performance of 3D CNNs is the increased computational complexity requiring large scale annotated datasets to train them in scale. 3D kernel factorization approaches have been proposed to reduce the complexity of 3D CNNs. Existing kernel factorization approaches follow hand-designed and hard-wired techniques. In this paper we propose Gate-Shift-Fuse (GSF), a novel spatio-temporal feature extraction module which controls interactions in spatio-temporal decomposition and learns to adaptively route features through time and combine them in a data dependent manner. GSF leverages grouped spatial gating to decompose input tensor and channel weighting to fuse the decomposed tensors. GSF can be inserted into existing 2D CNNs to convert them into an efficient and high performing spatio-temporal feature extractor, with negligible parameter and compute overhead. We perform an extensive analysis of GSF using two popular 2D CNN families and achieve state-of-the-art or competitive performance on five standard action recognition benchmarks.
While video action recognition has been an active area of research for several years, zero-shot action recognition has only recently started gaining traction. In this work, we propose a novel end-to-end trained transformer model which is capable of capturing long range spatiotemporal dependencies efficiently, contrary to existing approaches which use 3D-CNNs. Moreover, to address a common ambiguity in the existing works about classes that can be considered as previously unseen, we propose a new experimentation setup that satisfies the zero-shot learning premise for action recognition by avoiding overlap between the training and testing classes. The proposed approach significantly outperforms the state of the arts in zero-shot action recognition in terms of the the top-1 accuracy on UCF-101, HMDB-51 and ActivityNet datasets. The code and proposed experimentation setup are available in GitHub: https://github.com/Secure-and-Intelligent-Systems-Lab/SemanticVideoTransformer
Temporal Action Localization (TAL) has experienced remarkable success under the supervised learning paradigm. However, existing TAL methods are rooted in the closed set assumption, which cannot handle the inevitable unknown actions in open-world scenarios. In this paper, we, for the first time, step toward the Open Set TAL (OSTAL) problem and propose a general framework OpenTAL based on Evidential Deep Learning (EDL). Specifically, the OpenTAL consists of uncertainty-aware action classification, actionness prediction, and temporal location regression. With the proposed importance-balanced EDL method, classification uncertainty is learned by collecting categorical evidence majorly from important samples. To distinguish the unknown actions from background video frames, the actionness is learned by the positive-unlabeled learning. The classification uncertainty is further calibrated by leveraging the guidance from the temporal localization quality. The OpenTAL is general to enable existing TAL models for open set scenarios, and experimental results on THUMOS14 and ActivityNet1.3 benchmarks show the effectiveness of our method. The code and pre-trained models are released at https://www.rit.edu/actionlab/opental.
Weakly supervised temporal action localization aims to localize temporal boundaries of actions and simultaneously identify their categories with only video-level category labels. Many existing methods seek to generate pseudo labels for bridging the discrepancy between classification and localization, but usually only make use of limited contextual information for pseudo label generation. To alleviate this problem, we propose a representative snippet summarization and propagation framework. Our method seeks to mine the representative snippets in each video for propagating information between video snippets to generate better pseudo labels. For each video, its own representative snippets and the representative snippets from a memory bank are propagated to update the input features in an intra- and inter-video manner. The pseudo labels are generated from the temporal class activation maps of the updated features to rectify the predictions of the main branch. Our method obtains superior performance in comparison to the existing methods on two benchmarks, THUMOS14 and ActivityNet1.3, achieving gains as high as 1.2% in terms of average mAP on THUMOS14.
Making each modality in multi-modal data contribute is of vital importance to learning a versatile multi-modal model. Existing methods, however, are often dominated by one or few of modalities during model training, resulting in sub-optimal performance. In this paper, we refer to this problem as modality bias and attempt to study it in the context of multi-modal classification systematically and comprehensively. After stepping into several empirical analysis, we recognize that one modality affects the model prediction more just because this modality has a spurious correlation with instance labels. In order to primarily facilitate the evaluation on the modality bias problem, we construct two datasets respectively for the colored digit recognition and video action recognition tasks in line with the Out-of-Distribution (OoD) protocol. Collaborating with the benchmarks in the visual question answering task, we empirically justify the performance degradation of the existing methods on these OoD datasets, which serves as evidence to justify the modality bias learning. In addition, to overcome this problem, we propose a plug-and-play loss function method, whereby the feature space for each label is adaptively learned according to the training set statistics. Thereafter, we apply this method on eight baselines in total to test its effectiveness. From the results on four datasets regarding the above three tasks, our method yields remarkable performance improvements compared with the baselines, demonstrating its superiority on reducing the modality bias problem.
We consider the task of temporal human action localization in lifestyle vlogs. We introduce a novel dataset consisting of manual annotations of temporal localization for 13,000 narrated actions in 1,200 video clips. We present an extensive analysis of this data, which allows us to better understand how the language and visual modalities interact throughout the videos. We propose a simple yet effective method to localize the narrated actions based on their expected duration. Through several experiments and analyses, we show that our method brings complementary information with respect to previous methods, and leads to improvements over previous work for the task of temporal action localization.
Self-attention based Transformer models have demonstrated impressive results for image classification and object detection, and more recently for video understanding. Inspired by this success, we investigate the application of Transformer networks for temporal action localization in videos. To this end, we present ActionFormer -- a simple yet powerful model to identify actions in time and recognize their categories in a single shot, without using action proposals or relying on pre-defined anchor windows. ActionFormer combines a multiscale feature representation with local self-attention, and uses a light-weighted decoder to classify every moment in time and estimate the corresponding action boundaries. We show that this orchestrated design results in major improvements upon prior works. Without bells and whistles, ActionFormer achieves 71.0% mAP at tIoU=0.5 on THUMOS14, outperforming the best prior model by 14.1 absolute percentage points. Further, ActionFormer demonstrates strong results on ActivityNet 1.3 (36.6% average mAP) and EPIC-Kitchens 100 (+13.5% average mAP over prior works). Our code is available at http://github.com/happyharrycn/actionformer_release.
Human action recognition has been widely used in many fields of life, and many human action datasets have been published at the same time. However, most of the multi-modal databases have some shortcomings in the layout and number of sensors, which cannot fully represent the action features. Regarding the problems, this paper proposes a freely available dataset, named CZU-MHAD (Changzhou University: a comprehensive multi-modal human action dataset). It consists of 22 actions and three modals temporal synchronized data. These modals include depth videos and skeleton positions from a kinect v2 camera, and inertial signals from 10 wearable sensors. Compared with single modal sensors, multi-modal sensors can collect different modal data, so the use of multi-modal sensors can describe actions more accurately. Moreover, CZU-MHAD obtains the 3-axis acceleration and 3-axis angular velocity of 10 main motion joints by binding inertial sensors to them, and these data were captured at the same time. Experimental results are provided to show that this dataset can be used to study structural relationships between different parts of the human body when performing actions and fusion approaches that involve multi-modal sensor data.
Human action recognition still exists many challenging problems such as different viewpoints, occlusion, lighting conditions, human body size and the speed of action execution, although it has been widely used in different areas. To tackle these challenges, the Kinect depth sensor has been developed to record real time depth sequences, which are insensitive to the color of human clothes and illumination conditions. Many methods on recognizing human action have been reported in the literature such as HON4D, HOPC, RBD and HDG, which use the 4D surface normals, pointclouds, skeleton-based model and depth gradients respectively to capture discriminative information from depth videos or skeleton data. In this research project, the performance of four aforementioned algorithms will be analyzed and evaluated using five benchmark datasets, which cover challenging issues such as noise, change of viewpoints, background clutters and occlusions. We also implemented and improved the HDG algorithm, and applied it in cross-view action recognition using the UWA3D Multiview Activity dataset. Moreover, we used different combinations of individual feature vectors in HDG for performance evaluation. The experimental results show that our improvement of HDG outperforms other three state-of-the-art algorithms for cross-view action recognition.
Temporal action proposal generation (TAPG) is a challenging task that aims to locate action instances in untrimmed videos with temporal boundaries. To evaluate the confidence of proposals, the existing works typically predict action score of proposals that are supervised by the temporal Intersection-over-Union (tIoU) between proposal and the ground-truth. In this paper, we innovatively propose a general auxiliary Background Constraint idea to further suppress low-quality proposals, by utilizing the background prediction score to restrict the confidence of proposals. In this way, the Background Constraint concept can be easily plug-and-played into existing TAPG methods (e.g., BMN, GTAD). From this perspective, we propose the Background Constraint Network (BCNet) to further take advantage of the rich information of action and background. Specifically, we introduce an Action-Background Interaction module for reliable confidence evaluation, which models the inconsistency between action and background by attention mechanisms at the frame and clip levels. Extensive experiments are conducted on two popular benchmarks, i.e., ActivityNet-1.3 and THUMOS14. The results demonstrate that our method outperforms state-of-the-art methods. Equipped with the existing action classifier, our method also achieves remarkable performance on the temporal action localization task.
Recently, video-based action recognition methods using convolutional neural networks (CNNs) achieve remarkable recognition performance. However, there is still lack of understanding about the generalization mechanism of action recognition models. In this paper, we suggest that action recognition models rely on the motion information less than expected, and thus they are robust to randomization of frame orders. Furthermore, we find that motion monotonicity remaining after randomization also contributes to such robustness. Based on this observation, we develop a novel defense method using temporal shuffling of input videos against adversarial attacks for action recognition models. Another observation enabling our defense method is that adversarial perturbations on videos are sensitive to temporal destruction. To the best of our knowledge, this is the first attempt to design a defense method without additional training for 3D CNN-based video action recognition models.
Modern self-supervised learning algorithms typically enforce persistency of instance representations across views. While being very effective on learning holistic image and video representations, such an objective becomes sub-optimal for learning spatio-temporally fine-grained features in videos, where scenes and instances evolve through space and time. In this paper, we present Contextualized Spatio-Temporal Contrastive Learning (ConST-CL) to effectively learn spatio-temporally fine-grained video representations via self-supervision. We first design a region-based pretext task which requires the model to transform in-stance representations from one view to another, guided by context features. Further, we introduce a simple network design that successfully reconciles the simultaneous learning process of both holistic and local representations. We evaluate our learned representations on a variety of downstream tasks and show that ConST-CL achieves competitive results on 6 datasets, including Kinetics, UCF, HMDB, AVA-Kinetics, AVA and OTB.
Temporal action detection aims to locate the boundaries of action in the video. The current method based on boundary matching enumerates and calculates all possible boundary matchings to generate proposals. However, these methods neglect the long-range context aggregation in boundary prediction. At the same time, due to the similar semantics of adjacent matchings, local semantic aggregation of densely-generated matchings cannot improve semantic richness and discrimination. In this paper, we propose the end-to-end proposal generation method named Dual Context Aggregation Network (DCAN) to aggregate context on two levels, namely, boundary level and proposal level, for generating high-quality action proposals, thereby improving the performance of temporal action detection. Specifically, we design the Multi-Path Temporal Context Aggregation (MTCA) to achieve smooth context aggregation on boundary level and precise evaluation of boundaries. For matching evaluation, Coarse-to-fine Matching (CFM) is designed to aggregate context on the proposal level and refine the matching map from coarse to fine. We conduct extensive experiments on ActivityNet v1.3 and THUMOS-14. DCAN obtains an average mAP of 35.39% on ActivityNet v1.3 and reaches mAP 54.14% at [email omitted] on THUMOS-14, which demonstrates DCAN can generate high-quality proposals and achieve state-of-the-art performance. We release the code at https://github.com/cg1177/DCAN.
Semi-supervised video action recognition tends to enable deep neural networks to achieve remarkable performance even with very limited labeled data. However, existing methods are mainly transferred from current image-based methods (e.g., FixMatch). Without specifically utilizing the temporal dynamics and inherent multimodal attributes, their results could be suboptimal. To better leverage the encoded temporal information in videos, we introduce temporal gradient as an additional modality for more attentive feature extraction in this paper. To be specific, our method explicitly distills the fine-grained motion representations from temporal gradient (TG) and imposes consistency across different modalities (i.e., RGB and TG). The performance of semi-supervised action recognition is significantly improved without additional computation or parameters during inference. Our method achieves the state-of-the-art performance on three video action recognition benchmarks (i.e., Kinetics-400, UCF-101, and HMDB-51) under several typical semi-supervised settings (i.e., different ratios of labeled data).