In this paper we discuss several forms of spatiotemporal convolutions for video analysis and study their effects on action recognition. Our motivation stems from the observation that 2D CNNs applied to individual frames of the video have remained solid performers in action recognition. In this work we empirically demonstrate the accuracy advantages of 3D CNNs over 2D CNNs within the framework of residual learning. Furthermore, we show that factorizing the 3D convolutional filters into separate spatial and temporal components yields significantly advantages in accuracy. Our empirical study leads to the design of a new spatiotemporal convolutional block "R(2+1)D" which gives rise to CNNs that achieve results comparable or superior to the state-of-the-art on Sports-1M, Kinetics, UCF101 and HMDB51.
Locating actions in long untrimmed videos has been a challenging problem in video content analysis. The performances of existing action localization approaches remain unsatisfactory in precisely determining the beginning and the end of an action. Imitating the human perception procedure with observations and refinements, we propose a novel three-phase action localization framework. Our framework is embedded with an Actionness Network to generate initial proposals through frame-wise similarity grouping, and then a Refinement Network to conduct boundary adjustment on these proposals. Finally, the refined proposals are sent to a Localization Network for further fine-grained location regression. The whole process can be deemed as multi-stage refinement using a novel non-local pyramid feature under various temporal granularities. We evaluate our framework on THUMOS14 benchmark and obtain a significant improvement over the state-of-the-arts approaches. Specifically, the performance gain is remarkable under precise localization with high IoU thresholds. Our proposed framework achieves mAP@IoU=0.5 of 34.2%.
Automatically generating a summary of sports video poses the challenge of detecting interesting moments, or highlights, of a game. Traditional sports video summarization methods leverage editing conventions of broadcast sports video that facilitate the extraction of high-level semantics. However, user-generated videos are not edited, and thus traditional methods are not suitable to generate a summary. In order to solve this problem, this work proposes a novel video summarization method that uses players' actions as a cue to determine the highlights of the original video. A deep neural network-based approach is used to extract two types of action-related features and to classify video segments into interesting or uninteresting parts. The proposed method can be applied to any sports in which games consist of a succession of actions. Especially, this work considers the case of Kendo (Japanese fencing) as an example of a sport to evaluate the proposed method. The method is trained using Kendo videos with ground truth labels that indicate the video highlights. The labels are provided by annotators possessing different experience with respect to Kendo to demonstrate how the proposed method adapts to different needs. The performance of the proposed method is compared with several combinations of different features, and the results show that it outperforms previous summarization methods.
A new large-scale video dataset for human action recognition, called STAIR Actions is introduced. STAIR Actions contains 100 categories of action labels representing fine-grained everyday home actions so that it can be applied to research in various home tasks such as nursing, caring, and security. In STAIR Actions, each video has a single action label. Moreover, for each action category, there are around 1,000 videos that were obtained from YouTube or produced by crowdsource workers. The duration of each video is mostly five to six seconds. The total number of videos is 102,462. We explain how we constructed STAIR Actions and show the characteristics of STAIR Actions compared to existing datasets for human action recognition. Experiments with three major models for action recognition show that STAIR Actions can train large models and achieve good performance. STAIR Actions can be downloaded from http://actions.stair.center
Human skeleton joints are popular for action analysis since they can be easily extracted from videos to discard background noises. However, current skeleton representations do not fully benefit from machine learning with CNNs. We propose "Skepxels" a spatio-temporal representation for skeleton sequences to fully exploit the "local" correlations between joints using the 2D convolution kernels of CNN. We transform skeleton videos into images of flexible dimensions using Skepxels and develop a CNN-based framework for effective human action recognition using the resulting images. Skepxels encode rich spatio-temporal information about the skeleton joints in the frames by maximizing a unique distance metric, defined collaboratively over the distinct joint arrangements used in the skeletal image. Moreover, they are flexible in encoding compound semantic notions such as location and speed of the joints. The proposed action recognition exploits the representation in a hierarchical manner by first capturing the micro-temporal relations between the skeleton joints with the Skepxels and then exploiting their macro-temporal relations by computing the Fourier Temporal Pyramids over the CNN features of the skeletal images. We extend the Inception-ResNet CNN architecture with the proposed method and improve the state-of-the-art accuracy by 4.4% on the large scale NTU human activity dataset. On the medium-sized N-UCLA and UTH-MHAD datasets, our method outperforms the existing results by 5.7% and 9.3% respectively.
Data of different modalities generally convey complimentary but heterogeneous information, and a more discriminative representation is often preferred by combining multiple data modalities like the RGB and infrared features. However in reality, obtaining both data channels is challenging due to many limitations. For example, the RGB surveillance cameras are often restricted from private spaces, which is in conflict with the need of abnormal activity detection for personal security. As a result, using partial data channels to build a full representation of multi-modalities is clearly desired. In this paper, we propose a novel Partial-modal Generative Adversarial Networks (PM-GANs) that learns a full-modal representation using data from only partial modalities. The full representation is achieved by a generated representation in place of the missing data channel. Extensive experiments are conducted to verify the performance of our proposed method on action recognition, compared with four state-of-the-art methods. Meanwhile, a new Infrared-Visible Dataset for action recognition is introduced, and will be the first publicly available action dataset that contains paired infrared and visible spectrum.
Skeleton-based human action recognition has recently drawn increasing attentions with the availability of large-scale skeleton datasets. The most crucial factors for this task lie in two aspects: the intra-frame representation for joint co-occurrences and the inter-frame representation for skeletons' temporal evolutions. In this paper we propose an end-to-end convolutional co-occurrence feature learning framework. The co-occurrence features are learned with a hierarchical methodology, in which different levels of contextual information are aggregated gradually. Firstly point-level information of each joint is encoded independently. Then they are assembled into semantic representation in both spatial and temporal domains. Specifically, we introduce a global spatial aggregation scheme, which is able to learn superior joint co-occurrence features over local aggregation. Besides, raw skeleton coordinates as well as their temporal difference are integrated with a two-stream paradigm. Experiments show that our approach consistently outperforms other state-of-the-arts on action recognition and detection benchmarks like NTU RGB+D, SBU Kinect Interaction and PKU-MMD.
This paper proposes three simple, compact yet effective representations of depth sequences, referred to respectively as Dynamic Depth Images (DDI), Dynamic Depth Normal Images (DDNI) and Dynamic Depth Motion Normal Images (DDMNI), for both isolated and continuous action recognition. These dynamic images are constructed from a segmented sequence of depth maps using hierarchical bidirectional rank pooling to effectively capture the spatial-temporal information. Specifically, DDI exploits the dynamics of postures over time and DDNI and DDMNI exploit the 3D structural information captured by depth maps. Upon the proposed representations, a ConvNet based method is developed for action recognition. The image-based representations enable us to fine-tune the existing Convolutional Neural Network (ConvNet) models trained on image data without training a large number of parameters from scratch. The proposed method achieved the state-of-art results on three large datasets, namely, the Large-scale Continuous Gesture Recognition Dataset (means Jaccard index 0.4109), the Large-scale Isolated Gesture Recognition Dataset (59.21%), and the NTU RGB+D Dataset (87.08% cross-subject and 84.22% cross-view) even though only the depth modality was used.
We propose TAL-Net, an improved approach to temporal action localization in video that is inspired by the Faster R-CNN object detection framework. TAL-Net addresses three key shortcomings of existing approaches: (1) we improve receptive field alignment using a multi-scale architecture that can accommodate extreme variation in action durations; (2) we better exploit the temporal context of actions for both proposal generation and action classification by appropriately extending receptive fields; and (3) we explicitly consider multi-stream feature fusion and demonstrate that fusing motion late is important. We achieve state-of-the-art performance for both action proposal and localization on THUMOS'14 detection benchmark and competitive performance on ActivityNet challenge.
Skeleton-based human action recognition has recently attracted increasing attention thanks to the accessibility and the popularity of 3D skeleton data. One of the key challenges in skeleton-based action recognition lies in the large view variations when capturing data. In order to alleviate the effects of view variations, this paper introduces a novel view adaptation scheme, which automatically determines the virtual observation viewpoints in a learning based data driven manner. We design two view adaptive neural networks, i.e., VA-RNN based on RNN, and VA-CNN based on CNN. For each network, a novel view adaptation module learns and determines the most suitable observation viewpoints, and transforms the skeletons to those viewpoints for the end-to-end recognition with a main classification network. Ablation studies find that the proposed view adaptive models are capable of transforming the skeletons of various viewpoints to much more consistent virtual viewpoints which largely eliminates the viewpoint influence. In addition, we design a two-stream scheme (referred to as VA-fusion) that fuses the scores of the two networks to provide the fused prediction. Extensive experimental evaluations on five challenging benchmarks demonstrate that the effectiveness of the proposed view-adaptive networks and superior performance over state-of-the-art approaches. The source code is available at https://github.com/microsoft/View-Adaptive-Neural-Networks-for-Skeleton-based-Human-Action-Recognition.
From the frame/clip-level feature learning to the video-level representation building, deep learning methods in action recognition have developed rapidly in recent years. However, current methods suffer from the confusion caused by partial observation training, or without end-to-end learning, or restricted to single temporal scale modeling and so on. In this paper, we build upon two-stream ConvNets and propose Deep networks with Temporal Pyramid Pooling (DTPP), an end-to-end video-level representation learning approach, to address these problems. Specifically, at first, RGB images and optical flow stacks are sparsely sampled across the whole video. Then a temporal pyramid pooling layer is used to aggregate the frame-level features which consist of spatial and temporal cues. Lastly, the trained model has compact video-level representation with multiple temporal scales, which is both global and sequence-aware. Experimental results show that DTPP achieves the state-of-the-art performance on two challenging video action datasets: UCF101 and HMDB51, either by ImageNet pre-training or Kinetics pre-training.
Several theories in cognitive neuroscience suggest that when people interact with the world, or simulate interactions, they do so from a first-person egocentric perspective, and seamlessly transfer knowledge between third-person (observer) and first-person (actor). Despite this, learning such models for human action recognition has not been achievable due to the lack of data. This paper takes a step in this direction, with the introduction of Charades-Ego, a large-scale dataset of paired first-person and third-person videos, involving 112 people, with 4000 paired videos. This enables learning the link between the two, actor and observer perspectives. Thereby, we address one of the biggest bottlenecks facing egocentric vision research, providing a link from first-person to the abundant third-person data on the web. We use this data to learn a joint representation of first and third-person videos, with only weak supervision, and show its effectiveness for transferring knowledge from the third-person to the first-person domain.
Action recognition is a critical task for social robots to meaningfully engage with their environment. 3D human skeleton-based action recognition is an attractive research area in recent years. Although, the existing approaches are good at action recognition, it is a great challenge to recognize a group of actions in an activity scene. To tackle this problem, at first, we partition the scene into several primitive actions (PAs) based upon motion attention mechanism. Then, the primitive actions are described by the trajectory vectors of corresponding joints. After that, motivated by text classification based on word embedding, we employ convolution neural network (CNN) to recognize activity scenes by considering motion of joints as "word" of activity. The experimental results on the scenes of human activity dataset show the efficiency of the proposed approach.
We propose Human Pose Models that represent RGB and depth images of human poses independent of clothing textures, backgrounds, lighting conditions, body shapes and camera viewpoints. Learning such universal models requires training images where all factors are varied for every human pose. Capturing such data is prohibitively expensive. Therefore, we develop a framework for synthesizing the training data. First, we learn representative human poses from a large corpus of real motion captured human skeleton data. Next, we fit synthetic 3D humans with different body shapes to each pose and render each from 180 camera viewpoints while randomly varying the clothing textures, background and lighting. Generative Adversarial Networks are employed to minimize the gap between synthetic and real image distributions. CNN models are then learned that transfer human poses to a shared high-level invariant space. The learned CNN models are then used as invariant feature extractors from real RGB and depth frames of human action videos and the temporal variations are modelled by Fourier Temporal Pyramid. Finally, linear SVM is used for classification. Experiments on three benchmark cross-view human action datasets show that our algorithm outperforms existing methods by significant margins for RGB only and RGB-D action recognition.
This paper introduces a video dataset of spatio-temporally localized Atomic Visual Actions (AVA). The AVA dataset densely annotates 80 atomic visual actions in 430 15-minute video clips, where actions are localized in space and time, resulting in 1.58M action labels with multiple labels per person occurring frequently. The key characteristics of our dataset are: (1) the definition of atomic visual actions, rather than composite actions; (2) precise spatio-temporal annotations with possibly multiple annotations for each person; (3) exhaustive annotation of these atomic actions over 15-minute video clips; (4) people temporally linked across consecutive segments; and (5) using movies to gather a varied set of action representations. This departs from existing datasets for spatio-temporal action recognition, which typically provide sparse annotations for composite actions in short video clips. We will release the dataset publicly. AVA, with its realistic scene and action complexity, exposes the intrinsic difficulty of action recognition. To benchmark this, we present a novel approach for action localization that builds upon the current state-of-the-art methods, and demonstrates better performance on JHMDB and UCF101-24 categories. While setting a new state of the art on existing datasets, the overall results on AVA are low at 15.6% mAP, underscoring the need for developing new approaches for video understanding.
Skeleton-based action recognition task is entangled with complex spatio-temporal variations of skeleton joints, and remains challenging for Recurrent Neural Networks (RNNs). In this work, we propose a temporal-then-spatial recalibration scheme to alleviate such complex variations, resulting in an end-to-end Memory Attention Networks (MANs) which consist of a Temporal Attention Recalibration Module (TARM) and a Spatio-Temporal Convolution Module (STCM). Specifically, the TARM is deployed in a residual learning module that employs a novel attention learning network to recalibrate the temporal attention of frames in a skeleton sequence. The STCM treats the attention calibrated skeleton joint sequences as images and leverages the Convolution Neural Networks (CNNs) to further model the spatial and temporal information of skeleton data. These two modules (TARM and STCM) seamlessly form a single network architecture that can be trained in an end-to-end fashion. MANs significantly boost the performance of skeleton-based action recognition and achieve the best results on four challenging benchmark datasets: NTU RGB+D, HDM05, SYSU-3D and UT-Kinect.
Automatic analysis of the video is one of most complex problems in the fields of computer vision and machine learning. A significant part of this research deals with (human) activity recognition (HAR) since humans, and the activities that they perform, generate most of the video semantics. Video-based HAR has applications in various domains, but one of the most important and challenging is HAR in sports videos. Some of the major issues include high inter- and intra-class variations, large class imbalance, the presence of both group actions and single player actions, and recognizing simultaneous actions, i.e., the multi-label learning problem. Keeping in mind these challenges and the recent success of CNNs in solving various computer vision problems, in this work, we implement a 3D CNN based multi-label deep HAR system for multi-label class-imbalanced action recognition in hockey videos. We test our system for two different scenarios: an ensemble of $k$ binary networks vs. a single $k$-output network, on a publicly available dataset. We also compare our results with the system that was originally designed for the chosen dataset. Experimental results show that the proposed approach performs better than the existing solution.
Spatiotemporal feature learning in videos is a fundamental problem in computer vision. This paper presents a new architecture, termed as Appearance-and-Relation Network (ARTNet), to learn video representation in an end-to-end manner. ARTNets are constructed by stacking multiple generic building blocks, called as SMART, whose goal is to simultaneously model appearance and relation from RGB input in a separate and explicit manner. Specifically, SMART blocks decouple the spatiotemporal learning module into an appearance branch for spatial modeling and a relation branch for temporal modeling. The appearance branch is implemented based on the linear combination of pixels or filter responses in each frame, while the relation branch is designed based on the multiplicative interactions between pixels or filter responses across multiple frames. We perform experiments on three action recognition benchmarks: Kinetics, UCF101, and HMDB51, demonstrating that SMART blocks obtain an evident improvement over 3D convolutions for spatiotemporal feature learning. Under the same training setting, ARTNets achieve superior performance on these three datasets to the existing state-of-the-art methods.
Despite the significant progress that has been made on estimating optical flow recently, most estimation methods, including classical and deep learning approaches, still have difficulty with multi-scale estimation, real-time computation, and/or occlusion reasoning. In this paper, we introduce dilated convolution and occlusion reasoning into unsupervised optical flow estimation to address these issues. The dilated convolution allows our network to avoid upsampling via deconvolution and the resulting gridding artifacts. Dilated convolution also results in a smaller memory footprint which speeds up interference. The occlusion reasoning prevents our network from learning incorrect deformations due to occluded image regions during training. Our proposed method outperforms state-of-the-art unsupervised approaches on the KITTI benchmark. We also demonstrate its generalization capability by applying it to action recognition in video.
With the fast development of effective and low-cost human skeleton capture systems, skeleton-based action recognition has attracted much attention recently. Most existing methods use Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) to extract spatio-temporal information embedded in the skeleton sequences for action recognition. However, these approaches are limited in the ability of relational modeling in a single skeleton, due to the loss of important structural information when converting the raw skeleton data to adapt to the input format of CNN or RNN. In this paper, we propose an Attentional Recurrent Relational Network-LSTM (ARRN-LSTM) to simultaneously model spatial configurations and temporal dynamics in skeletons for action recognition. We introduce the Recurrent Relational Network to learn the spatial features in a single skeleton, followed by a multi-layer LSTM to learn the temporal features in the skeleton sequences. Between the two modules, we design an adaptive attentional module to focus attention on the most discriminative parts in the single skeleton. To exploit the complementarity from different geometries in the skeleton for sufficient relational modeling, we design a two-stream architecture to learn the structural features among joints and lines simultaneously. Extensive experiments are conducted on several popular skeleton datasets and the results show that the proposed approach achieves better results than most mainstream methods.