Recently, tensor data (or multidimensional array) have been generated in many modern applications, such as functional magnetic resonance imaging (fMRI) in neuroscience and videos in video analysis. Many efforts are made in recent years to predict the relationship between tensor features and univariate responses. However, previously proposed methods either lose structural information within tensor data or have prohibitively expensive time costs, especially for large-scale data with high-order structures. To address such problems, we propose the Sparse and Low-rank Tensor Regression (SLTR) model. Our model enforces sparsity and low-rankness of the tensor coefficient by directly applying $\ell_1$ norm and tensor nuclear norm, such that it preserves structural information of the tensor. To make the solving procedure scalable and efficient, SLTR makes use of the proximal gradient method, which can be easily implemented parallelly. We evaluate SLTR on several simulated datasets and one video action recognition dataset. Experiment results show that, compared with previous models, SLTR can obtain a better solution with much fewer time costs. Moreover, our model's predictions exhibit meaningful interpretations on the video dataset.
Temporal action detection is a fundamental yet challenging task in video understanding. Video context is a critical cue to effectively detect actions, but current works mainly focus on temporal context, while neglecting semantic context as well as other important context properties. In this work, we propose a graph convolutional network (GCN) model to adaptively incorporate multi-level semantic context into video features and cast temporal action detection as a sub-graph localization problem. Specifically, we formulate video snippets as graph nodes, snippet-snippet correlations as edges, and actions associated with context as target sub-graphs. With graph convolution as the basic operation, we design a GCN block called GCNeXt, which learns the features of each node by aggregating its context and dynamically updates the edges in the graph. To localize each sub-graph, we also design an SGAlign layer to embed each sub-graph into the Euclidean space. Extensive experiments show that G-TAD is capable of finding effective video context without extra supervision and achieves state-of-the-art performance on two detection benchmarks. On ActivityNet-1.3, it obtains an average mAP of 34.09%; on THUMOS14, it reaches 51.6% at [email omitted] when combined with a proposal processing method. G-TAD code is publicly available at https://github.com/frostinassiky/gtad.
How does one represent an action? How does one describe an action that we have never seen before? Such questions are addressed by the Zero Shot Learning paradigm, where a model is trained on only a subset of classes and is evaluated on its ability to correctly classify an example from a class it has never seen before. In this work, we present a body pose based zero shot action recognition network and demonstrate its performance on the NTU RGB-D dataset. Our model learns to jointly encapsulate visual similarities based on pose features of the action performer as well as similarities in the natural language descriptions of the unseen action class names. We demonstrate how this pose-language semantic space encodes knowledge which allows our model to correctly predict actions not seen during training.
Weakly-supervised temporal action localization is a very challenging problem because frame-wise labels are not given in the training stage while the only hint is video-level labels: whether each video contains action frames of interest. Previous methods aggregate frame-level class scores to produce video-level prediction and learn from video-level action labels. This formulation does not fully model the problem in that background frames are forced to be misclassified as action classes to predict video-level labels accurately. In this paper, we design Background Suppression Network (BaS-Net) which introduces an auxiliary class for background and has a two-branch weight-sharing architecture with an asymmetrical training strategy. This enables BaS-Net to suppress activations from background frames to improve localization performance. Extensive experiments demonstrate the effectiveness of BaS-Net and its superiority over the state-of-the-art methods on the most popular benchmarks - THUMOS'14 and ActivityNet. Our code and the trained model are available at https://github.com/Pilhyeon/BaSNet-pytorch.
Temporal localization remains an important challenge in video understanding. In this work, we present our solution to the 3rd YouTube-8M Video Understanding Challenge organized by Google Research. Participants were required to build a segment-level classifier using a large-scale training data set with noisy video-level labels and a relatively small-scale validation data set with accurate segment-level labels. We formulated the problem as a multiple instance multi-label learning and developed an attention-based mechanism to selectively emphasize the important frames by attention weights. The model performance is further improved by constructing multiple sets of attention networks. We further fine-tuned the model using the segment-level data set. Our final model consists of an ensemble of attention/multi-attention networks, deep bag of frames models, recurrent neural networks and convolutional neural networks. It ranked 13th on the private leader board and stands out for its efficient usage of resources.
Generating temporal action proposals remains a very challenging problem, where the main issue lies in predicting precise temporal proposal boundaries and reliable action confidence in long and untrimmed real-world videos. In this paper, we propose an efficient and unified framework to generate temporal action proposals named Dense Boundary Generator (DBG), which draws inspiration from boundary-sensitive methods and implements boundary classification and action completeness regression for densely distributed proposals. In particular, the DBG consists of two modules: Temporal boundary classification (TBC) and Action-aware completeness regression (ACR). The TBC aims to provide two temporal boundary confidence maps by low-level two-stream features, while the ACR is designed to generate an action completeness score map by high-level action-aware features. Moreover, we introduce a dual stream BaseNet (DSB) to encode RGB and optical flow information, which helps to capture discriminative boundary and actionness features. Extensive experiments on popular benchmarks ActivityNet-1.3 and THUMOS14 demonstrate the superiority of DBG over the state-of-the-art proposal generator (e.g., MGG and BMN). Our code will be made available upon publication.
Temporal action localization in untrimmed videos is an important but difficult task. Difficulties are encountered in the application of existing methods when modeling temporal structures of videos. In the present study, we developed a novel method, referred to as Gemini Network, for effective modeling of temporal structures and achieving high-performance temporal action localization. The significant improvements afforded by the proposed method are attributable to three major factors. First, the developed network utilizes two subnets for effective modeling of temporal structures. Second, three parallel feature extraction pipelines are used to prevent interference between the extractions of different stage features. Third, the proposed method utilizes auxiliary supervision, with the auxiliary classifier losses affording additional constraints for improving the modeling capability of the network. As a demonstration of its effectiveness, the Gemini Network was used to achieve state-of-the-art temporal action localization performance on two challenging datasets, namely, THUMOS14 and ActivityNet.
Multimodal fusion frameworks for Human Action Recognition (HAR) using depth and inertial sensor data have been proposed over the years. In most of the existing works, fusion is performed at a single level (feature level or decision level), missing the opportunity to fuse rich mid-level features necessary for better classification. To address this shortcoming, in this paper, we propose three novel deep multilevel multimodal fusion frameworks to capitalize on different fusion strategies at various stages and to leverage the superiority of multilevel fusion. At input, we transform the depth data into depth images called sequential front view images (SFIs) and inertial sensor data into signal images. Each input modality, depth and inertial, is further made multimodal by taking convolution with the Prewitt filter. Creating "modality within modality" enables further complementary and discriminative feature extraction through Convolutional Neural Networks (CNNs). CNNs are trained on input images of each modality to learn low-level, high-level and complex features. Learned features are extracted and fused at different stages of the proposed frameworks to combine discriminative and complementary information. These highly informative features are served as input to a multi-class Support Vector Machine (SVM). We evaluate the proposed frameworks on three publicly available multimodal HAR datasets, namely, UTD Multimodal Human Action Dataset (MHAD), Berkeley MHAD, and UTD-MHAD Kinect V2. Experimental results show the supremacy of the proposed fusion frameworks over existing methods.
Recently, Weakly-supervised Temporal Action Localization (WTAL) has been densely studied but there is still a large gap between weakly-supervised models and fully-supervised models. It is practical and intuitive to annotate temporal boundaries of a few examples and utilize them to help WTAL models better detect actions. However, the train-test discrepancy of action localization strategy prevents WTAL models from leveraging semi-supervision for further improvement. At training time, attention or multiple instance learning is used to aggregate predictions of each snippet for video-level classification; at test time, they first obtain action score sequences over time, then truncate segments of scores higher than a fixed threshold, and post-process action segments. The inconsistent strategy makes it hard to explicitly supervise the action localization model with temporal boundary annotations at training time. In this paper, we propose a Train-Test Consistent framework, TTC-Loc. In both training and testing time, our TTC-Loc localizes actions by comparing scores of action classes and predicted threshold, which enables it to be trained with semi-supervision. By fixing the train-test discrepancy, our TTC-Loc significantly outperforms the state-of-the-art performance on THUMOS'14, ActivityNet 1.2 and 1.3 when only video-level labels are provided for training. With full annotations of only one video per class and video-level labels for the other videos, our TTC-Loc further boosts the performance and achieves 33.4\% mAP (IoU threshold 0.5) on THUMOS's 14.
Drones are enabling new forms of human actions surveillance due to their low cost and fast mobility. However, using deep neural networks for automatic aerial action recognition is difficult due to the need for a large number of training aerial human action videos. Collecting a large number of human action aerial videos is costly, time-consuming, and difficult. In this paper, we explore two alternative data sources to improve aerial action classification when only a few training aerial examples are available. As a first data source, we resort to video games. We collect plenty of aerial game action videos using two gaming engines. For the second data source, we leverage conditional Wasserstein Generative Adversarial Networks to generate aerial features from ground videos. Given that both data sources have some limitations, e.g. game videos are biased towards specific actions categories (fighting, shooting, etc.,), and it is not easy to generate good discriminative GAN-generated features for all types of actions, we need to efficiently integrate two dataset sources with few available real aerial training videos. To address this challenge of the heterogeneous nature of the data, we propose to use a disjoint multitask learning framework. We feed the network with real and game, or real and GAN-generated data in an alternating fashion to obtain an improved action classifier. We validate the proposed approach on two aerial action datasets and demonstrate that features from aerial game videos and those generated from GAN can be extremely useful for an improved action recognition in real aerial videos when only a few real aerial training examples are available.
Convolutional neural networks are nowadays witnessing a major success in different pattern recognition problems. These learning models were basically designed to handle vectorial data such as images but their extension to non-vectorial and semi-structured data (namely graphs with variable sizes, topology, etc.) remains a major challenge, though a few interesting solutions are currently emerging. In this paper, we introduce MLGCN; a novel spectral Multi-Laplacian Graph Convolutional Network. The main contribution of this method resides in a new design principle that learns graph-laplacians as convex combinations of other elementary laplacians each one dedicated to a particular topology of the input graphs. We also introduce a novel pooling operator, on graphs, that proceeds in two steps: context-dependent node expansion is achieved, followed by a global average pooling; the strength of this two-step process resides in its ability to preserve the discrimination power of nodes while achieving permutation invariance. Experiments conducted on SBU and UCF-101 datasets, show the validity of our method for the challenging task of action recognition.
Deep video action recognition models have been highly successful in recent years but require large quantities of manually annotated data, which are expensive and laborious to obtain. In this work, we investigate the generation of synthetic training data for video action recognition, as synthetic data have been successfully used to supervise models for a variety of other computer vision tasks. We propose an interpretable parametric generative model of human action videos that relies on procedural generation, physics models and other components of modern game engines. With this model we generate a diverse, realistic, and physically plausible dataset of human action videos, called PHAV for "Procedural Human Action Videos". PHAV contains a total of 39,982 videos, with more than 1,000 examples for each of 35 action categories. Our video generation approach is not limited to existing motion capture sequences: 14 of these 35 categories are procedurally defined synthetic actions. In addition, each video is represented with 6 different data modalities, including RGB, optical flow and pixel-level semantic labels. These modalities are generated almost simultaneously using the Multiple Render Targets feature of modern GPUs. In order to leverage PHAV, we introduce a deep multi-task (i.e. that considers action classes from multiple datasets) representation learning architecture that is able to simultaneously learn from synthetic and real video datasets, even when their action categories differ. Our experiments on the UCF-101 and HMDB-51 benchmarks suggest that combining our large set of synthetic videos with small real-world datasets can boost recognition performance. Our approach also significantly outperforms video representations produced by fine-tuning state-of-the-art unsupervised generative models of videos.
3D skeleton-based action recognition and motion prediction are two essential problems of human activity understanding. In many previous works: 1) they studied two tasks separately, neglecting internal correlations; 2) they did not capture sufficient relations inside the body. To address these issues, we propose a symbiotic model to handle two tasks jointly; and we propose two scales of graphs to explicitly capture relations among body-joints and body-parts. Together, we propose symbiotic graph neural networks, which contain a backbone, an action-recognition head, and a motion-prediction head. Two heads are trained jointly and enhance each other. For the backbone, we propose multi-branch multi-scale graph convolution networks to extract spatial and temporal features. The multi-scale graph convolution networks are based on joint-scale and part-scale graphs. The joint-scale graphs contain actional graphs, capturing action-based relations, and structural graphs, capturing physical constraints. The part-scale graphs integrate body-joints to form specific parts, representing high-level relations. Moreover, dual bone-based graphs and networks are proposed to learn complementary features. We conduct extensive experiments for skeleton-based action recognition and motion prediction with four datasets, NTU-RGB+D, Kinetics, Human3.6M, and CMU Mocap. Experiments show that our symbiotic graph neural networks achieve better performances on both tasks compared to the state-of-the-art methods.
Effective processing of video input is essential for the recognition of temporally varying events such as human actions. Motivated by the often distinctive temporal characteristics of actions in either horizontal or vertical direction, we introduce a novel convolution block for CNN architectures with video input. Our proposed Fractioned Adjacent Spatial and Temporal (FAST) 3D convolutions are a natural decomposition of a regular 3D convolution. Each convolution block consist of three sequential convolution operations: a 2D spatial convolution followed by spatio-temporal convolutions in the horizontal and vertical direction, respectively. Additionally, we introduce a FAST variant that treats horizontal and vertical motion in parallel. Experiments on benchmark action recognition datasets UCF-101 and HMDB-51 with ResNet architectures demonstrate consistent increased performance of FAST 3D convolution blocks over traditional 3D convolutions. The lower validation loss indicates better generalization, especially for deeper networks. We also evaluate the performance of CNN architectures with similar memory requirements, based either on Two-stream networks or with 3D convolution blocks. DenseNet-121 with FAST 3D convolutions was shown to perform best, giving further evidence of the merits of the decoupled spatio-temporal convolutions.
Automatic detection of individual intake gestures during eating occasions has the potential to improve dietary monitoring and support dietary recommendations. Existing studies typically make use of on-body solutions such as inertial and audio sensors, while video is used as ground truth. Intake gesture detection directly based on video has rarely been attempted. In this study, we address this gap and show that deep learning architectures can successfully be applied to the problem of video-based detection of intake gestures. For this purpose, we collect and label video data of eating occasions using 360-degree video of 102 participants. Applying state-of-the-art approaches from video action recognition, our results show that (1) the best model achieves an $F_1$ score of 0.858, (2) appearance features contribute more than motion features, and (3) temporal context in form of multiple video frames is essential for top model performance.
In the few-shot scenario, a learner must effectively generalize to unseen classes given a small support set of labeled examples. While a relatively large amount of research has gone into few-shot learning for image classification, little work has been done on few-shot video classification. In this work, we address the task of few-shot video action recognition with a set of two-stream models. We evaluate the performance of a set of convolutional and recurrent neural network video encoder architectures used in conjunction with three popular metric-based few-shot algorithms. We train and evaluate using a few-shot split of the Kinetics 600 dataset. Our experiments confirm the importance of the two-stream setup, and find prototypical networks and pooled long short-term memory network embeddings to give the best performance as few-shot method and video encoder, respectively. For a 5-shot 5-way task, this setup obtains 84.2% accuracy on the test set and 59.4% on a special "challenge" test set, composed of highly confusable classes.
Deep convolutional networks are widely used in video action recognition. 3D convolutions are one prominent approach to deal with the additional time dimension. While 3D convolutions typically lead to higher accuracies, the inner workings of the trained models are more difficult to interpret. We focus on creating human-understandable visual explanations that represent the hierarchical parts of spatio-temporal networks. We introduce Class Feature Pyramids, a method that traverses the entire network structure and incrementally discovers kernels at different network depths that are informative for a specific class. Our method does not depend on the network's architecture or the type of 3D convolutions, supporting grouped and depth-wise convolutions, convolutions in fibers, and convolutions in branches. We demonstrate the method on six state-of-the-art 3D convolution neural networks (CNNs) on three action recognition (Kinetics-400, UCF-101, and HMDB-51) and two egocentric action recognition datasets (EPIC-Kitchens and EGTEA Gaze+).
Recently, infrared human action recognition has attracted increasing attention for it has many advantages over visible light, that is, being robust to illumination change and shadows. However, the infrared action data is limited until now, which degrades the performance of infrared action recognition. Motivated by the idea of transfer learning, an infrared human action recognition framework using auxiliary data from visible light is proposed to solve the problem of limited infrared action data. In the proposed framework, we first construct a novel Cross-Dataset Feature Alignment and Generalization (CDFAG) framework to map the infrared data and visible light data into a common feature space, where Kernel Manifold Alignment (KEMA) and a dual alignedto-generalized encoders (AGE) model are employed to represent the feature. Then, a support vector machine (SVM) is trained, using both the infrared data and visible light data, and can classify the features derived from infrared data. The proposed method is evaluated on InfAR, which is a publicly available infrared human action dataset. To build up auxiliary data, we set up a novel visible light action dataset XD145. Experimental results show that the proposed method can achieve state-of-the-art performance compared with several transfer learning and domain adaptation methods.
Infrared human action recognition has many advantages, i.e., it is insensitive to illumination change, appearance variability, and shadows. Existing methods for infrared action recognition are either based on spatial or local temporal information, however, the global temporal information, which can better describe the movements of body parts across the whole video, is not considered. In this letter, we propose a novel global temporal representation named optical-flow stacked difference image (OFSDI) and extract robust and discriminative feature from the infrared action data by considering the local, global, and spatial temporal information together. Due to the small size of the infrared action dataset, we first apply convolutional neural networks on local, spatial, and global temporal stream respectively to obtain efficient convolutional feature maps from the raw data rather than train a classifier directly. Then these convolutional feature maps are aggregated into effective descriptors named three-stream trajectory-pooled deep-convolutional descriptors by trajectory-constrained pooling. Furthermore, we improve the robustness of these features by using the locality-constrained linear coding (LLC) method. With these features, a linear support vector machine (SVM) is adopted to classify the action data in our scheme. We conduct the experiments on infrared action recognition datasets InfAR and NTU RGB+D. The experimental results show that the proposed approach outperforms the representative state-of-the-art handcrafted features and deep learning features based methods for the infrared action recognition.
Deep learning models achieve impressive performance for skeleton-based human action recognition. However, the robustness of these models to adversarial attacks remains largely unexplored due to their complex spatio-temporal nature that must represent sparse and discrete skeleton joints. This work presents the first adversarial attack on skeleton-based action recognition with graph convolutional networks. The proposed targeted attack, termed Constrained Iterative Attack for Skeleton Actions (CIASA), perturbs joint locations in an action sequence such that the resulting adversarial sequence preserves the temporal coherence, spatial integrity, and the anthropomorphic plausibility of the skeletons. CIASA achieves this feat by satisfying multiple physical constraints, and employing spatial skeleton realignments for the perturbed skeletons along with regularization of the adversarial skeletons with Generative networks. We also explore the possibility of semantically imperceptible localized attacks with CIASA, and succeed in fooling the state-of-the-art skeleton action recognition models with high confidence. CIASA perturbations show high transferability for black-box attacks. We also show that the perturbed skeleton sequences are able to induce adversarial behavior in the RGB videos created with computer graphics. A comprehensive evaluation with NTU and Kinetics datasets ascertains the effectiveness of CIASA for graph-based skeleton action recognition and reveals the imminent threat to the spatio-temporal deep learning tasks in general.