Weakly supervised temporal action localization is a challenging vision task due to the absence of ground-truth temporal locations of actions in the training videos. With only video-level supervision during training, most existing methods rely on a Multiple Instance Learning (MIL) framework to predict the start and end frame of each action category in a video. However, the existing MIL-based approach has a major limitation of only capturing the most discriminative frames of an action, ignoring the full extent of an activity. Moreover, these methods cannot model background activity effectively, which plays an important role in localizing foreground activities. In this paper, we present a novel framework named HAM-Net with a hybrid attention mechanism which includes temporal soft, semi-soft and hard attentions to address these issues. Our temporal soft attention module, guided by an auxiliary background class in the classification module, models the background activity by introducing an "action-ness" score for each video snippet. Moreover, our temporal semi-soft and hard attention modules, calculating two attention scores for each video snippet, help to focus on the less discriminative frames of an action to capture the full action boundary. Our proposed approach outperforms recent state-of-the-art methods by at least 2.2% mAP at IoU threshold 0.5 on the THUMOS14 dataset, and by at least 1.3% mAP at IoU threshold 0.75 on the ActivityNet1.2 dataset. Code can be found at: https://github.com/asrafulashiq/hamnet.
Current developments in temporal event or action localization usually target actions captured by a single camera. However, extensive events or actions in the wild may be captured as a sequence of shots by multiple cameras at different positions. In this paper, we propose a new and challenging task called multi-shot temporal event localization, and accordingly, collect a large scale dataset called MUlti-Shot EventS (MUSES). MUSES has 31,477 event instances for a total of 716 video hours. The core nature of MUSES is the frequent shot cuts, for an average of 19 shots per instance and 176 shots per video, which induces large intrainstance variations. Our comprehensive evaluations show that the state-of-the-art method in temporal action localization only achieves an mAP of 13.1% at IoU=0.5. As a minor contribution, we present a simple baseline approach for handling the intra-instance variations, which reports an mAP of 18.9% on MUSES and 56.9% on THUMOS14 at IoU=0.5. To facilitate research in this direction, we release the dataset and the project code at https://songbai.site/muses/ .
3D Convolutional Neural Network (3D CNN) captures spatial and temporal information on 3D data such as video sequences. However, due to the convolution and pooling mechanism, the information loss seems unavoidable. To improve the visual explanations and classification in 3D CNN, we propose two approaches; i) aggregate layer-wise global to local (global-local) discrete gradients using trained 3DResNext network, and ii) implement attention gating network to improve the accuracy of the action recognition. The proposed approach intends to show the usefulness of every layer termed as global-local attention in 3D CNN via visual attribution, weakly-supervised action localization, and action recognition. Firstly, the 3DResNext is trained and applied for action classification using backpropagation concerning the maximum predicted class. The gradients and activations of every layer are then up-sampled. Later, aggregation is used to produce more nuanced attention, which points out the most critical part of the predicted class's input videos. We use contour thresholding of final attention for final localization. We evaluate spatial and temporal action localization in trimmed videos using fine-grained visual explanation via 3DCam. Experimental results show that the proposed approach produces informative visual explanations and discriminative attention. Furthermore, the action recognition via attention gating on each layer produces better classification results than the baseline model.
Point-Level temporal action localization (PTAL) aims to localize actions in untrimmed videos with only one timestamp annotation for each action instance. Existing methods adopt the frame-level prediction paradigm to learn from the sparse single-frame labels. However, such a framework inevitably suffers from a large solution space. This paper attempts to explore the proposal-based prediction paradigm for point-level annotations, which has the advantage of more constrained solution space and consistent predictions among neighboring frames. The point-level annotations are first used as the keypoint supervision to train a keypoint detector. At the location prediction stage, a simple but effective mapper module, which enables back-propagation of training errors, is then introduced to bridge the fully-supervised framework with weak supervision. To our best of knowledge, this is the first work to leverage the fully-supervised paradigm for the point-level setting. Experiments on THUMOS14, BEOID, and GTEA verify the effectiveness of our proposed method both quantitatively and qualitatively, and demonstrate that our method outperforms state-of-the-art methods.
Spatiotemporal action recognition deals with locating and classifying actions in videos. Motivated by the latest state-of-the-art real-time object detector You Only Watch Once (YOWO), we aim to modify its structure to increase action detection precision and reduce computational time. Specifically, we propose four novel approaches in attempts to improve YOWO and address the imbalanced class issue in videos by modifying the loss function. We consider two moderate-sized datasets to apply our modification of YOWO - the popular Joint-annotated Human Motion Data Base (J-HMDB-21) and a private dataset of restaurant video footage provided by a Carnegie Mellon University-based startup, Agot.AI. The latter involves fast-moving actions with small objects as well as unbalanced data classes, making the task of action localization more challenging. We implement our proposed methods in the GitHub repository https://github.com/stoneMo/YOWOv2.
This work proposes a weakly-supervised temporal action localization framework, called D2-Net, which strives to temporally localize actions using video-level supervision. Our main contribution is the introduction of a novel loss formulation, which jointly enhances the discriminability of latent embeddings and robustness of the output temporal class activations with respect to foreground-background noise caused by weak supervision. The proposed formulation comprises a discriminative and a denoising loss term for enhancing temporal action localization. The discriminative term incorporates a classification loss and utilizes a top-down attention mechanism to enhance the separability of latent foreground-background embeddings. The denoising loss term explicitly addresses the foreground-background noise in class activations by simultaneously maximizing intra-video and inter-video mutual information using a bottom-up attention mechanism. As a result, activations in the foreground regions are emphasized whereas those in the background regions are suppressed, thereby leading to more robust predictions. Comprehensive experiments are performed on multiple benchmarks, including THUMOS14 and ActivityNet1.2. Our D2-Net performs favorably in comparison to the existing methods on all datasets, achieving gains as high as 2.3% in terms of mAP at IoU=0.5 on THUMOS14. Source code is available at https://github.com/naraysa/D2-Net
State-of-the-art video action classifiers often suffer from overfitting. They tend to be biased towards specific objects and scene cues, rather than the foreground action content, leading to sub-optimal generalization performances. Recent data augmentation strategies have been reported to address the overfitting problems in static image classifiers. Despite the effectiveness on the static image classifiers, data augmentation has rarely been studied for videos. For the first time in the field, we systematically analyze the efficacy of various data augmentation strategies on the video classification task. We then propose a powerful augmentation strategy VideoMix. VideoMix creates a new training video by inserting a video cuboid into another video. The ground truth labels are mixed proportionally to the number of voxels from each video. We show that VideoMix lets a model learn beyond the object and scene biases and extract more robust cues for action recognition. VideoMix consistently outperforms other augmentation baselines on Kinetics and the challenging Something-Something-V2 benchmarks. It also improves the weakly-supervised action localization performance on THUMOS'14. VideoMix pretrained models exhibit improved accuracies on the video detection task (AVA).
Temporal action localization (TAL) in videos is a challenging task, especially due to the large variation in action temporal scales. Short actions usually occupy a major proportion in the datasets, but tend to have the lowest performance. In this paper, we confront the challenge of short actions and propose a multi-level cross-scale solution dubbed as video self-stitching graph network (VSGN). We have two key components in VSGN: video self-stitching (VSS) and cross-scale graph pyramid network (xGPN). In VSS, we focus on a short period of a video and magnify it along the temporal dimension to obtain a larger scale. We stitch the original clip and its magnified counterpart in one input sequence to take advantage of the complementary properties of both scales. The xGPN component further exploits the cross-scale correlations by a pyramid of cross-scale graph networks, each containing a hybrid module to aggregate features from across scales as well as within the same scale. Our VSGN not only enhances the feature representations, but also generates more positive anchors for short actions and more short training samples. Experiments demonstrate that VSGN obviously improves the localization performance of short actions as well as achieving the state-of-the-art overall performance on THUMOS-14 and ActivityNet-v1.3.
Due to the large memory footprint of untrimmed videos, current state-of-the-art video localization methods operate atop precomputed video clip features. These features are extracted from video encoders typically trained for trimmed action classification tasks, making such features not necessarily suitable for temporal localization. In this work, we propose a novel supervised pretraining paradigm for clip features that not only trains to classify activities but also considers background clips and global video information to improve temporal sensitivity. Extensive experiments show that using features trained with our novel pretraining strategy significantly improves the performance of recent state-of-the-art methods on three tasks: Temporal Action Localization, Action Proposal Generation, and Dense Video Captioning. We also show that our pretraining approach is effective across three encoder architectures and two pretraining datasets. We believe video feature encoding is an important building block for localization algorithms, and extracting temporally-sensitive features should be of paramount importance in building more accurate models. The code and pretrained models are available on our project website.
Literature on self-assessment in machine learning mainly focuses on the production of well-calibrated algorithms through consensus frameworks i.e. calibration is seen as a problem. Yet, we observe that learning to be properly confident could behave like a powerful regularization and thus, could be an opportunity to improve performance.Precisely, we show that used within a framework of action detection, the learning of a self-assessment score is able to improve the whole action localization process.Experimental results show that our approach outperforms the state-of-the-art on two action detection benchmarks. On THUMOS14 dataset, the mAP at [email omitted] is improved from 42.8\% to 44.6\%, and from 50.4\% to 51.7\% on ActivityNet1.3 dataset. For lower tIoU values, we achieve even more significant improvements on both datasets.
Weakly-supervised Temporal Action Localization (W-TAL) aims to classify and localize all action instances in an untrimmed video under only video-level supervision. However, without frame-level annotations, it is challenging for W-TAL methods to identify false positive action proposals and generate action proposals with precise temporal boundaries. In this paper, we present a Two-Stream Consensus Network (TSCN) to simultaneously address these challenges. The proposed TSCN features an iterative refinement training method, where a frame-level pseudo ground truth is iteratively updated, and used to provide frame-level supervision for improved model training and false positive action proposal elimination. Furthermore, we propose a new attention normalization loss to encourage the predicted attention to act like a binary selection, and promote the precise localization of action instance boundaries. Experiments conducted on the THUMOS14 and ActivityNet datasets show that the proposed TSCN outperforms current state-of-the-art methods, and even achieves comparable results with some recent fully-supervised methods.
Spatio-temporal action localization is an important problem in computer vision that involves detecting where and when activities occur, and therefore requires modeling of both spatial and temporal features. This problem is typically formulated in the context of supervised learning, where the learned classifiers operate on the premise that both training and test data are sampled from the same underlying distribution. However, this assumption does not hold when there is a significant domain shift, leading to poor generalization performance on the test data. To address this, we focus on the hard and novel task of generalizing training models to test samples without access to any labels from the latter for spatio-temporal action localization by proposing an end-to-end unsupervised domain adaptation algorithm. We extend the state-of-the-art object detection framework to localize and classify actions. In order to minimize the domain shift, three domain adaptation modules at image level (temporal and spatial) and instance level (temporal) are designed and integrated. We design a new experimental setup and evaluate the proposed method and different adaptation modules on the UCF-Sports, UCF-101 and JHMDB benchmark datasets. We show that significant performance gain can be achieved when spatial and temporal features are adapted separately, or jointly for the most effective results.
Generating human action proposals in untrimmed videos is an important yet challenging task with wide applications. Current methods often suffer from the noisy boundary locations and the inferior quality of confidence scores used for proposal retrieving. In this paper, we present BSN++, a new framework which exploits complementary boundary regressor and relation modeling for temporal proposal generation. First, we propose a novel boundary regressor based on the complementary characteristics of both starting and ending boundary classifiers. Specifically, we utilize the U-shaped architecture with nested skip connections to capture rich contexts and introduce bi-directional boundary matching mechanism to improve boundary precision. Second, to account for the proposal-proposal relations ignored in previous methods, we devise a proposal relation block to which includes two self-attention modules from the aspects of position and channel. Furthermore, we find that there inevitably exists data imbalanced problems in the positive/negative proposals and temporal durations, which harm the model performance on tail distributions. To relieve this issue, we introduce the scale-balanced re-sampling strategy. Extensive experiments are conducted on two popular benchmarks: ActivityNet-1.3 and THUMOS14, which demonstrate that BSN++ achieves the state-of-the-art performance. Not surprisingly, the proposed BSN++ ranked 1st place in the CVPR19 - ActivityNet challenge leaderboard on temporal action localization task.
With the knowledge of action moments (i.e., trimmed video clips that each contains an action instance), humans could routinely localize an action temporally in an untrimmed video. Nevertheless, most practical methods still require all training videos to be labeled with temporal annotations (action category and temporal boundary) and develop the models in a fully-supervised manner, despite expensive labeling efforts and inapplicable to new categories. In this paper, we introduce a new design of transfer learning type to learn action localization for a large set of action categories, but only on action moments from the categories of interest and temporal annotations of untrimmed videos from a small set of action classes. Specifically, we present Action Herald Networks (AherNet) that integrate such design into an one-stage action localization framework. Technically, a weight transfer function is uniquely devised to build the transformation between classification of action moments or foreground video segments and action localization in synthetic contextual moments or untrimmed videos. The context of each moment is learnt through the adversarial mechanism to differentiate the generated features from those of background in untrimmed videos. Extensive experiments are conducted on the learning both across the splits of ActivityNet v1.3 and from THUMOS14 to ActivityNet v1.3. Our AherNet demonstrates the superiority even comparing to most fully-supervised action localization methods. More remarkably, we train AherNet to localize actions from 600 categories on the leverage of action moments in Kinetics-600 and temporal annotations from 200 classes in ActivityNet v1.3. Source code and data are available at \url{https://github.com/FuchenUSTC/AherNet}.
This work addresses the problem of temporal action localization with Variance-Aware Networks (VAN), i.e., DNNs that use second-order statistics in the input and/or the output of regression tasks. We first propose a network (VANp) that when presented with the second-order statistics of the input, i.e., each sample has a mean and a variance, it propagates the mean and the variance throughout the network to deliver outputs with second order statistics. In this framework, both the input and the output could be interpreted as Gaussians. To do so, we derive differentiable analytic solutions, or reasonable approximations, to propagate across commonly used NN layers. To train the network, we define a differentiable loss based on the KL-divergence between the predicted Gaussian and a Gaussian around the ground truth action borders, and use standard back-propagation. Importantly, the variances propagation in VANp does not require any additional parameters, and during testing, does not require any additional computations either. In action localization, the means and the variances of the input are computed at pooling operations, that are typically used to bring arbitrarily long videos to a vector with fixed dimensions. Second, we propose two alternative formulations that augment the first (respectively, the last) layer of a regression network with additional parameters so as to take in the input (respectively, predict in the output) both means and variances. Results in the action localization problem show that the incorporation of second order statistics improves over the baseline network, and that VANp surpasses the accuracy of virtually all other two-stage networks without involving any additional parameters.
Most of the current action localization methods follow an anchor-based pipeline: depicting action instances by pre-defined anchors, learning to select the anchors closest to the ground truth, and predicting the confidence of anchors with refinements. Pre-defined anchors set prior about the location and duration for action instances, which facilitates the localization for common action instances but limits the flexibility for tackling action instances with drastic varieties, especially for extremely short or extremely long ones. To address this problem, this paper proposes a novel anchor-free action localization module that assists action localization by temporal points. Specifically, this module represents an action instance as a point with its distances to the starting boundary and ending boundary, alleviating the pre-defined anchor restrictions in terms of action localization and duration. The proposed anchor-free module is capable of predicting the action instances whose duration is either extremely short or extremely long. By combining the proposed anchor-free module with a conventional anchor-based module, we propose a novel action localization framework, called A2Net. The cooperation between anchor-free and anchor-based modules achieves superior performance to the state-of-the-art on THUMOS14 (45.5% vs. 42.8%). Furthermore, comprehensive experiments demonstrate the complementarity between the anchor-free and the anchor-based module, making A2Net simple but effective.
In this paper, we address the problem of temporal action localization with a single stage neural network. In the proposed architecture we model the boundary predictions as uni-variate Gaussian distributions in order to model their uncertainties, which is the first in this area to the best of our knowledge. We use two uncertainty-aware boundary regression losses: first, the Kullback-Leibler divergence between the ground truth location of the boundary and the Gaussian modeling the prediction of the boundary and second, the expectation of the $\ell_1$ loss under the same Gaussian. We show that with both uncertainty modeling approaches improve the detection performance by more than $1.5\%$ in mAP@tIoU=0.5 and that the proposed simple one-stage network performs closely to more complex one and two stage networks.
Most current pipelines for spatio-temporal action localization connect frame-wise or clip-wise detection results to generate action proposals, where only local information is exploited and the efficiency is hindered by dense per-frame localization. In this paper, we propose Coarse-to-Fine Action Detector (CFAD),an original end-to-end trainable framework for efficient spatio-temporal action localization. The CFAD introduces a new paradigm that first estimates coarse spatio-temporal action tubes from video streams, and then refines the tubes' location based on key timestamps. This concept is implemented by two key components, the Coarse and Refine Modules in our framework. The parameterized modeling of long temporal information in the Coarse Module helps obtain accurate initial tube estimation, while the Refine Module selectively adjusts the tube location under the guidance of key timestamps. Against other methods, theproposed CFAD achieves competitive results on action detection benchmarks of UCF101-24, UCFSports and JHMDB-21 with inference speed that is 3.3x faster than the nearest competitors.
Weakly supervised temporal action localization is a newly emerging yet widely studied topic in recent years. The existing methods can be categorized into two localization-by-classification pipelines, i.e., the pre-classification pipeline and the post-classification pipeline. The pre-classification pipeline first performs classification on each video snippet and then aggregate the snippet-level classification scores to obtain the video-level classification score. In contrast, the post-classification pipeline aggregates the snippet-level features first and then predicts the video-level classification score based on the aggregated feature. Although the classifiers in these two pipelines are used in different ways, the role they play is exactly the same---to classify the given features to identify the corresponding action categories. To this end, an ideal classifier can make both pipelines work. This inspires us to simultaneously learn these two pipelines in a unified framework to obtain an effective classifier. Specifically, in the proposed learning framework, we implement two parallel network streams to model the two localization-by-classification pipelines simultaneously and make the two network streams share the same classifier. This achieves the novel Equivalent Classification Mapping (ECM) mechanism. Moreover, we discover that an ideal classifier may possess two characteristics: 1) The frame-level classification scores obtained from the pre-classification stream and the feature aggregation weights in the post-classification stream should be consistent; 2) The classification results of these two streams should be identical. Based on these two characteristics, we further introduce a weight-transition module and an equivalent training strategy into the proposed learning framework, which assists to thoroughly mine the equivalence mechanism.
This paper strives to localize the temporal extent of an action in a long untrimmed video. Where existing work leverages many examples with their start, their ending, and/or the class of the action during training time, we propose few-shot common action localization. The start and end of an action in a long untrimmed video is determined based on just a hand-full of trimmed video examples containing the same action, without knowing their common class label. To address this task, we introduce a new 3D convolutional network architecture able to align representations from the support videos with the relevant query video segments. The network contains: (\textit{i}) a mutual enhancement module to simultaneously complement the representation of the few trimmed support videos and the untrimmed query video; (\textit{ii}) a progressive alignment module that iteratively fuses the support videos into the query branch; and (\textit{iii}) a pairwise matching module to weigh the importance of different support videos. Evaluation of few-shot common action localization in untrimmed videos containing a single or multiple action instances demonstrates the effectiveness and general applicability of our proposal.