Learning text-video embeddings usually requires a dataset of video clips with manually provided captions. However, such datasets are expensive and time consuming to create and therefore difficult to obtain on a large scale. In this work, we propose instead to learn such embeddings from video data with readily available natural language annotations in the form of automatically transcribed narrations. The contributions of this work are three-fold. First, we introduce HowTo100M: a large-scale dataset of 136 million video clips sourced from 1.22M narrated instructional web videos depicting humans performing and describing over 23k different visual tasks. Our data collection procedure is fast, scalable and does not require any additional manual annotation. Second, we demonstrate that a text-video embedding trained on this data leads to state-of-the-art results for text-to-video retrieval and action localization on instructional video datasets such as YouCook2 or CrossTask. Finally, we show that this embedding transfers well to other domains: fine-tuning on generic Youtube videos (MSR-VTT dataset) and movies (LSMDC dataset) outperforms models trained on these datasets alone. Our dataset, code and models will be publicly available at: www.di.ens.fr/willow/research/howto100m/.
Spatio-temporal action localization consists of three levels of tasks: spatial localization, action classification, and temporal segmentation. In this work, we propose a new Progressive Cross-stream Cooperation (PCSC) framework to use both region proposals and features from one stream (i.e. Flow/RGB) to help another stream (i.e. RGB/Flow) to iteratively improve action localization results and generate better bounding boxes in an iterative fashion. Specifically, we first generate a larger set of region proposals by combining the latest region proposals from both streams, from which we can readily obtain a larger set of labelled training samples to help learn better action detection models. Second, we also propose a new message passing approach to pass information from one stream to another stream in order to learn better representations, which also leads to better action detection models. As a result, our iterative framework progressively improves action localization results at the frame level. To improve action localization results at the video level, we additionally propose a new strategy to train class-specific actionness detectors for better temporal segmentation, which can be readily learnt by focusing on "confusing" samples from the same action class. Comprehensive experiments on two benchmark datasets UCF-101-24 and J-HMDB demonstrate the effectiveness of our newly proposed approaches for spatio-temporal action localization in realistic scenarios.
Temporal action localization has recently attracted significant interest in the Computer Vision community. However, despite the great progress, it is hard to identify which aspects of the proposed methods contribute most to the increase in localization performance. To address this issue, we conduct ablative experiments on feature extraction methods, fixed-size feature representation methods and training strategies, and report how each influences the overall performance. Based on our findings, we propose a two-stage detector that outperforms the state of the art in THUMOS14, achieving a mAP@tIoU=0.5 equal to 44.2%.
In weakly-supervised temporal action localization, previous works have failed to locate dense and integral regions for each entire action due to the overestimation of the most salient regions. To alleviate this issue, we propose a marginalized average attentional network (MAAN) to suppress the dominant response of the most salient regions in a principled manner. The MAAN employs a novel marginalized average aggregation (MAA) module and learns a set of latent discriminative probabilities in an end-to-end fashion. MAA samples multiple subsets from the video snippet features according to a set of latent discriminative probabilities and takes the expectation over all the averaged subset features. Theoretically, we prove that the MAA module with learned latent discriminative probabilities successfully reduces the difference in responses between the most salient regions and the others. Therefore, MAAN is able to generate better class activation sequences and identify dense and integral action regions in the videos. Moreover, we propose a fast algorithm to reduce the complexity of constructing MAA from O($2^T$) to O($T^2$). Extensive experiments on two large-scale video datasets show that our MAAN achieves superior performance on weakly-supervised temporal action localization
Events in natural videos typically arise from spatio-temporal interactions between actors and objects and involve multiple co-occurring activities and object classes. To capture this rich visual and semantic context, we propose using two graphs: (1) an attributed spatio-temporal visual graph whose nodes correspond to actors and objects and whose edges encode different types of interactions, and (2) a symbolic graph that models semantic relationships. We further propose a graph neural network for refining the representations of actors, objects and their interactions on the resulting hybrid graph. Our model goes beyond current approaches that assume nodes and edges are of the same type, operate on graphs with fixed edge weights and do not use a symbolic graph. In particular, our framework: a) has specialized attention-based message functions for different node and edge types; b) uses visual edge features; c) integrates visual evidence with label relationships; and d) performs global reasoning in the semantic space. Experiments on challenging video understanding tasks, such as temporal action localization on the Charades dataset, show that the proposed method leads to state-of-the-art performance.
Enabling computational systems with the ability to localize actions in video-based content has manifold applications. Traditionally, such a problem is approached in a fully-supervised setting where video-clips with complete frame-by-frame annotations around the actions of interest are provided for training. However, the data requirements needed to achieve adequate generalization in this setting is prohibitive. In this work, we circumvent this issue by casting the problem in a weakly supervised setting, i.e., by considering videos as labelled `sets' of unlabelled video segments. Firstly, we apply unsupervised segmentation to take advantage of the elementary structure of each video. Subsequently, a convolutional neural network is used to extract RGB features from the resulting video segments. Finally, Multiple Instance Learning (MIL) is employed to predict labels at the video segment level, thus inherently performing spatio-temporal action detection. In contrast to previous work, we make use of a different MIL formulation in which the label of each video segment is continuous rather then discrete, making the resulting optimization function tractable. Additionally, we utilize a set splitting technique for regularization. Experimental results considering multiple performance indicators on the UCF-Sports data-set support the effectiveness of our approach.
Human doing actions will result in WiFi distortion, which is widely explored for action recognition, such as the elderly fallen detection, hand sign language recognition, and keystroke estimation. As our best survey, past work recognizes human action by categorizing one complete distortion series into one action, which we term as series-level action recognition. In this paper, we introduce a much more fine-grained and challenging action recognition task into WiFi sensing domain, i.e., sample-level action recognition. In this task, every WiFi distortion sample in the whole series should be categorized into one action, which is a critical technique in precise action localization, continuous action segmentation, and real-time action recognition. To achieve WiFi-based sample-level action recognition, we fully analyze approaches in image-based semantic segmentation as well as in video-based frame-level action recognition, then propose a simple yet efficient deep convolutional neural network, i.e., Temporal Unet. Experimental results show that Temporal Unet achieves this novel task well. Codes have been made publicly available at https://github.com/geekfeiw/WiSLAR.
Detecting temporal extents of human actions in videos is a challenging computer vision problem that requires detailed manual supervision including frame-level labels. This expensive annotation process limits deploying action detectors to a limited number of categories. We propose a novel method, called WSGN, that learns to detect actions from \emph{weak supervision}, using only video-level labels. WSGN learns to exploit both video-specific and dataset-wide statistics to predict relevance of each frame to an action category. This strategy leads to significant gains in action detection for two standard benchmarks THUMOS14 and Charades. Our method obtains excellent results compared to state-of-the-art methods that uses similar features and loss functions on THUMOS14 dataset. Similarly, our weakly supervised method is only 0.3% mAP behind a state-of-the-art supervised method on challenging Charades dataset for action localization.
Video action detectors are usually trained using datasets with fully-supervised temporal annotations. Building such datasets is an expensive task. To alleviate this problem, recent methods have tried to leverage weak labeling, where videos are untrimmed and only a video-level label is available. In this paper, we propose RefineLoc, a novel weakly-supervised temporal action localization method. RefineLoc uses an iterative refinement approach by estimating and training on snippet-level pseudo ground truth at every iteration. We show the benefit of this iterative approach and present an extensive analysis of five different pseudo ground truth generators. We show the effectiveness of our model on two standard action datasets, ActivityNet v1.2 and THUMOS14. RefineLoc shows competitive results with the state-of-the-art in weakly-supervised temporal localization. Additionally, our iterative refinement process is able to significantly improve the performance of two state-of-the-art methods, setting a new state-of-the-art on THUMOS14.
Previous spatial-temporal action localization methods commonly follow the pipeline of object detection to estimate bounding boxes and labels of actions. However, the temporal relation of an action has not been fully explored. In this paper, we propose an end-to-end Progress Regression Recurrent Neural Network (PR-RNN) for online spatial-temporal action localization, which learns to infer the action by temporal progress regression. Two new action attributes, called progression and progress rate, are introduced to describe the temporal engagement and relative temporal position of an action. In our method, frame-level features are first extracted by a Fully Convolutional Network (FCN). Subsequently, detection results and action progress attributes are regressed by the Convolutional Gated Recurrent Unit (ConvGRU) based on all the observed frames instead of a single frame or a short clip. Finally, a novel online linking method is designed to connect single-frame results to spatial-temporal tubes with the help of the estimated action progress attributes. Extensive experiments demonstrate that the progress attributes improve the localization accuracy by providing more precise temporal position of an action in unconstrained videos. Our proposed PR-RNN achieves the stateof-the-art performance for most of the IoU thresholds on two benchmark datasets.
Temporal action localization is an important task of computer vision. Though many methods have been proposed, it still remains an open question how to predict the temporal location of action segments precisely. Most state-of-the-art works train action classifiers on video segments pre-determined by action proposal. However, recent work found that a desirable model should move beyond segment-level and make dense predictions at a fine granularity in time to determine precise temporal boundaries. In this paper, we propose a Frame Segmentation Network (FSN) that places a temporal CNN on top of the 2D spatial CNNs. Spatial CNNs are responsible for abstracting semantics in spatial dimension while temporal CNN is responsible for introducing temporal context information and performing dense predictions. The proposed FSN can make dense predictions at frame-level for a video clip using both spatial and temporal context information. FSN is trained in an end-to-end manner, so the model can be optimized in spatial and temporal domain jointly. We also adapt FSN to use it in weakly supervised scenario (WFSN), where only video level labels are provided when training. Experiment results on public dataset show that FSN achieves superior performance in both frame-level action localization and temporal action localization.
The task of action recognition or action detection involves analyzing videos and determining what action or motion is being performed. The primary subject of these videos are predominantly humans performing some action. However, this requirement can be relaxed to generalize over other subjects such as animals or robots. The applications can range from anywhere between human-computer inter-action to automated video editing proposals. When we consider spatiotemporal action recognition, we deal with action localization. This task not only involves determining what action is being performed but also when and where itis being performed in said video. This paper aims to survey the plethora of approaches and algorithms attempted to solve this task, give a comprehensive comparison between them, explore various datasets available for the problem, and determine the most promising approaches.
In this paper, we deal with the problem of temporal action localization for a large-scale untrimmed cricket videos dataset. Our action of interest for cricket videos is a cricket stroke played by a batsman, which is, usually, covered by cameras placed at the stands of the cricket ground at both ends of the cricket pitch. After applying a sequence of preprocessing steps, we have ~73 million frames for 1110 videos in the dataset at constant frame rate and resolution. The method of localization is a generalized one which applies a trained random forest model for CUTs detection(using summed up grayscale histogram difference features) and two linear SVM camera models(CAM1 and CAM2) for first frame detection, trained on HOG features of CAM1 and CAM2 video shots. CAM1 and CAM2 are assumed to be part of the cricket stroke. At the predicted boundary positions, the HOG features of the first frames are computed and a simple algorithm was used to combine the positively predicted camera shots. In order to make the process as generic as possible, we did not consider any domain specific knowledge, such as tracking or specific shape and motion features. The detailed analysis of our methodology is provided along with the metrics used for evaluation of individual models, and the final predicted segments. We achieved a weighted mean TIoU of 0.5097 over a small sample of the test set.
Despite huge success in the image domain, modern detection models such as Faster R-CNN have not been used nearly as much for video analysis. This is arguably due to the fact that detection models are designed to operate on single frames and as a result do not have a mechanism for learning motion representations directly from video. We propose a learning procedure that allows detection models such as Faster R-CNN to learn motion features directly from the RGB video data while being optimized with respect to a pose estimation task. Given a pair of video frames---Frame A and Frame B---we force our model to predict human pose in Frame A using the features from Frame B. We do so by leveraging deformable convolutions across space and time. Our network learns to spatially sample features from Frame B in order to maximize pose detection accuracy in Frame A. This naturally encourages our network to learn motion offsets encoding the spatial correspondences between the two frames. We refer to these motion offsets as DiMoFs (Discriminative Motion Features). In our experiments we show that our training scheme helps learn effective motion cues, which can be used to estimate and localize salient human motion. Furthermore, we demonstrate that as a byproduct, our model also learns features that lead to improved pose detection in still-images, and better keypoint tracking. Finally, we show how to leverage our learned model for the tasks of spatiotemporal action localization and fine-grained action recognition.
In this paper, we propose an end-to-end capsule network for pixel level localization of actors and actions present in a video. The localization is performed based on a natural language query through which an actor and action are specified. We propose to encode both the video as well as textual input in the form of capsules, which provide more effective representation in comparison with standard convolution based features. We introduce a novel capsule based attention mechanism for fusion of video and text capsules for text selected video segmentation. The attention mechanism is performed via joint EM routing over video and text capsules for text selected actor and action localization. The existing works on actor-action localization are mainly focused on localization in a single frame instead of the full video. Different from existing works, we propose to perform the localization on all frames of the video. To validate the potential of the proposed network for actor and action localization on all the frames of a video, we extend an existing actor-action dataset (A2D) with annotations for all the frames. The experimental evaluation demonstrates the effectiveness of the proposed capsule network for text selective actor and action localization in videos, and it also improves upon the performance of the existing state-of-the art works on single frame-based localization.
Temporal segmentation of long videos is an important problem, that has largely been tackled through supervised learning, often requiring large amounts of annotated training data. In this paper, we tackle the problem of self-supervised temporal segmentation of long videos that alleviate the need for any supervision. We introduce a self-supervised, predictive learning framework that draws inspiration from cognitive psychology to segment long, visually complex videos into individual, stable segments that share the same semantics. We also introduce a new adaptive learning paradigm that helps reduce the effect of catastrophic forgetting in recurrent neural networks. Extensive experiments on three publicly available datasets - Breakfast Actions, 50 Salads, and INRIA Instructional Videos datasets show the efficacy of the proposed approach. We show that the proposed approach is able to outperform weakly-supervised and other unsupervised learning approaches by up to 24% and have competitive performance compared to fully supervised approaches. We also show that the proposed approach is able to learn highly discriminative features that help improve action recognition when used in a representation learning paradigm.
Despite tremendous progress achieved in temporal action detection, state-of-the-art methods still suffer from the sharp performance deterioration when localizing the starting and ending temporal action boundaries. Although most methods apply boundary regression paradigm to tackle this problem, we argue that the direct regression lacks detailed enough information to yield accurate temporal boundaries. In this paper, we propose a novel Boundary Likelihood Pinpointing (BLP) network to alleviate this deficiency of boundary regression and improve the localization accuracy. Given a loosely localized search interval that contains an action instance, BLP casts the problem of localizing temporal boundaries as that of assigning probabilities on each equally divided unit of this interval. These generated probabilities provide useful information regarding the boundary location of the action inside this search interval. Based on these probabilities, we introduce a boundary pinpointing paradigm to pinpoint the accurate boundaries under a simple probabilistic framework. Compared with other C3D feature based detectors, extensive experiments demonstrate that BLP significantly improves the localization performance of recent state-of-the-art detectors, and achieves competitive detection mAP on both THUMOS' 14 and ActivityNet datasets, particularly when the evaluation tIoU is high.
We propose 'Hide-and-Seek' a general purpose data augmentation technique, which is complementary to existing data augmentation techniques and is beneficial for various visual recognition tasks. The key idea is to hide patches in a training image randomly, in order to force the network to seek other relevant content when the most discriminative content is hidden. Our approach only needs to modify the input image and can work with any network to improve its performance. During testing, it does not need to hide any patches. The main advantage of Hide-and-Seek over existing data augmentation techniques is its ability to improve object localization accuracy in the weakly-supervised setting, and we therefore use this task to motivate the approach. However, Hide-and-Seek is not tied only to the image localization task, and can generalize to other forms of visual input like videos, as well as other recognition tasks like image classification, temporal action localization, semantic segmentation, emotion recognition, age/gender estimation, and person re-identification. We perform extensive experiments to showcase the advantage of Hide-and-Seek on these various visual recognition problems.
Weakly supervised temporal action localization, which aims at temporally locating action instances in untrimmed videos using only video-level class labels during training, is an important yet challenging problem in video analysis. Many current methods adopt the "localization by classification" framework: first do video classification, then locate temporal area contributing to the results most. However, this framework fails to locate the entire action instances and gives little consideration to the local context. In this paper, we present a novel architecture called Cascaded Pyramid Mining Network (CPMN) to address these issues using two effective modules. First, to discover the entire temporal interval of specific action, we design a two-stage cascaded module with proposed Online Adversarial Erasing (OAE) mechanism, where new and complementary regions are mined through feeding the erased feature maps of discovered regions back to the system. Second, to exploit hierarchical contextual information in videos and reduce missing detections, we design a pyramid module which produces a scale-invariant attention map through combining the feature maps from different levels. Final, we aggregate the results of two modules to perform action localization via locating high score areas in temporal Class Activation Sequence (CAS). Extensive experiments conducted on THUMOS14 and ActivityNet-1.3 datasets demonstrate the effectiveness of our method.
Current state-of-the-art approaches for spatio-temporal action localization rely on detections at the frame level and model temporal context with 3D ConvNets. Here, we go one step further and model spatio-temporal relations to capture the interactions between human actors, relevant objects and scene elements essential to differentiate similar human actions. Our approach is weakly supervised and mines the relevant elements automatically with an actor-centric relational network (ACRN). ACRN computes and accumulates pair-wise relation information from actor and global scene features, and generates relation features for action classification. It is implemented as neural networks and can be trained jointly with an existing action detection system. We show that ACRN outperforms alternative approaches which capture relation information, and that the proposed framework improves upon the state-of-the-art performance on JHMDB and AVA. A visualization of the learned relation features confirms that our approach is able to attend to the relevant relations for each action.