By extracting spatial and temporal characteristics in one network, the two-stream ConvNets can achieve the state-of-the-art performance in action recognition. However, such a framework typically suffers from the separately processing of spatial and temporal information between the two standalone streams and is hard to capture long-term temporal dependence of an action. More importantly, it is incapable of finding the salient portions of an action, say, the frames that are the most discriminative to identify the action. To address these problems, a \textbf{j}oint \textbf{n}etwork based \textbf{a}ttention (JNA) is proposed in this study. We find that the fully-connected fusion, branch selection and spatial attention mechanism are totally infeasible for action recognition. Thus in our joint network, the spatial and temporal branches share some information during the training stage. We also introduce an attention mechanism on the temporal domain to capture the long-term dependence meanwhile finding the salient portions. Extensive experiments are conducted on two benchmark datasets, UCF101 and HMDB51. Experimental results show that our method can improve the action recognition performance significantly and achieves the state-of-the-art results on both datasets.
In the scope of gestural action recognition, the size of the feature vector representing movements is in general quite large especially when full body movements are considered. Furthermore, this feature vector evolves during the movement performance so that a complete movement is fully represented by a matrix M of size DxT , whose element M i, j represents the value of feature i at timestamps j. Many studies have addressed dimensionality reduction considering only the size of the feature vector lying in R D to reduce both the variability of gestural sequences expressed in the reduced space, and the computational complexity of their processing. In return, very few of these methods have explicitly addressed the dimensionality reduction along the time axis. Yet this is a major issue when considering the use of elastic distances which are characterized by a quadratic complexity along the time axis. We present in this paper an evaluation of straightforward approaches aiming at reducing the dimensionality of the matrix M for each movement, leading to consider both the dimensionality reduction of the feature vector as well as its reduction along the time axis. The dimensionality reduction of the feature vector is achieved by selecting remarkable joints in the skeleton performing the movement, basically the extremities of the articulatory chains composing the skeleton. The temporal dimen-sionality reduction is achieved using either a regular or adaptive down-sampling that seeks to minimize the reconstruction error of the movements. Elastic and Euclidean kernels are then compared through support vector machine learning. Two data sets 1 that are widely referenced in the domain of human gesture recognition, and quite distinctive in terms of quality of motion capture, are used for the experimental assessment of the proposed approaches. On these data sets we experimentally show that it is feasible, and possibly desirable, to significantly reduce simultaneously the size of the feature vector and the number of skeleton frames to represent body movements while maintaining a very good recognition rate. The method proves to give satisfactory results at a level currently reached by state-of-the-art methods on these data sets. We experimentally show that the computational complexity reduction that is obtained makes this approach eligible for real-time applications.
Zero-Shot Learning (ZSL) promises to scale visual recognition by bypassing the conventional model training requirement of annotated examples for every category. This is achieved by establishing a mapping connecting low-level features and a semantic description of the label space, referred as visual-semantic mapping, on auxiliary data. Reusing the learned mapping to project target videos into an embedding space thus allows novel-classes to be recognised by nearest neighbour inference. However, existing ZSL methods suffer from auxiliary-target domain shift intrinsically induced by assuming the same mapping for the disjoint auxiliary and target classes. This compromises the generalisation accuracy of ZSL recognition on the target data. In this work, we improve the ability of ZSL to generalise across this domain shift in both model- and data-centric ways by formulating a visual-semantic mapping with better generalisation properties and a dynamic data re-weighting method to prioritise auxiliary data that are relevant to the target classes. Specifically: (1) We introduce a multi-task visual-semantic mapping to improve generalisation by constraining the semantic mapping parameters to lie on a low-dimensional manifold, (2) We explore prioritised data augmentation by expanding the pool of auxiliary data with additional instances weighted by relevance to the target domain. The proposed new model is applied to the challenging zero-shot action recognition problem to demonstrate its advantages over existing ZSL models.
The past few years have witnessed increasing research interest on covariance-based feature representation. A variety of methods have been proposed to boost its efficacy, with some recent ones resorting to nonlinear kernel technique. Noting that the essence of this feature representation is to characterise the underlying structure of visual features, this paper argues that an equally, if not more, important approach to boosting its efficacy shall be to improve the quality of this characterisation. Following this idea, we propose to exploit the structure sparsity of visual features in skeletal human action recognition, and compute sparse inverse covariance estimate (SICE) as feature representation. We discuss the advantage of this new representation on dealing with small sample, high dimensionality, and modelling capability. Furthermore, utilising the monotonicity property of SICE, we efficiently generate a hierarchy of SICE matrices to characterise the structure of visual features at different sparsity levels, and two discriminative learning algorithms are then developed to adaptively integrate them to perform recognition. As demonstrated by extensive experiments, the proposed representation leads to significantly improved recognition performance over the state-of-the-art comparable methods. In particular, as a method fully based on linear technique, it is comparable or even better than those employing nonlinear kernel technique. This result well demonstrates the value of exploiting structure sparsity for covariance-based feature representation.
Deep convolutional neural networks (CNNs) have proven highly effective for visual recognition, where learning a universal representation from activations of convolutional layer plays a fundamental problem. In this paper, we present Fisher Vector encoding with Variational Auto-Encoder (FV-VAE), a novel deep architecture that quantizes the local activations of convolutional layer in a deep generative model, by training them in an end-to-end manner. To incorporate FV encoding strategy into deep generative models, we introduce Variational Auto-Encoder model, which steers a variational inference and learning in a neural network which can be straightforwardly optimized using standard stochastic gradient method. Different from the FV characterized by conventional generative models (e.g., Gaussian Mixture Model) which parsimoniously fit a discrete mixture model to data distribution, the proposed FV-VAE is more flexible to represent the natural property of data for better generalization. Extensive experiments are conducted on three public datasets, i.e., UCF101, ActivityNet, and CUB-200-2011 in the context of video action recognition and fine-grained image classification, respectively. Superior results are reported when compared to state-of-the-art representations. Most remarkably, our proposed FV-VAE achieves to-date the best published accuracy of 94.2% on UCF101.
The number of categories for action recognition is growing rapidly and it has become increasingly hard to label sufficient training data for learning conventional models for all categories. Instead of collecting ever more data and labelling them exhaustively for all categories, an attractive alternative approach is zero-shot learning" (ZSL). To that end, in this study we construct a mapping between visual features and a semantic descriptor of each action category, allowing new categories to be recognised in the absence of any visual training data. Existing ZSL studies focus primarily on still images, and attribute-based semantic representations. In this work, we explore word-vectors as the shared semantic space to embed videos and category labels for ZSL action recognition. This is a more challenging problem than existing ZSL of still images and/or attributes, because the mapping between video spacetime features of actions and the semantic space is more complex and harder to learn for the purpose of generalising over any cross-category domain shift. To solve this generalisation problem in ZSL action recognition, we investigate a series of synergistic strategies to improve upon the standard ZSL pipeline. Most of these strategies are transductive in nature which means access to testing data in the training phase.
Within the field of image and video recognition, the traditional approach is a dataset split into fixed training and test partitions. However, the labelling of the training set is time-consuming, especially as datasets grow in size and complexity. Furthermore, this approach is not applicable to the home user, who wants to intuitively group their media without tirelessly labelling the content. Our interactive approach is able to iteratively cluster classes of images and video. Our approach is based around the concept of an image signature which, unlike a standard bag of words model, can express co-occurrence statistics as well as symbol frequency. We efficiently compute metric distances between signatures despite their inherent high dimensionality and provide discriminative feature selection, to allow common and distinctive elements to be identified from a small set of user labelled examples. These elements are then accentuated in the image signature to increase similarity between examples and pull correct classes together. By repeating this process in an online learning framework, the accuracy of similarity increases dramatically despite labelling only a few training examples. To demonstrate that the approach is agnostic to media type and features used, we evaluate on three image datasets (15 scene, Caltech101 and FG-NET), a mixed text and image dataset (ImageTag), a dataset used in active learning (Iris) and on three action recognition datasets (UCF11, KTH and Hollywood2). On the UCF11 video dataset, the accuracy is 86.7% despite using only 90 labelled examples from a dataset of over 1200 videos, instead of the standard 1122 training videos. The approach is both scalable and efficient, with a single iteration over the full UCF11 dataset of around 1200 videos taking approximately 1 minute on a standard desktop machine.
In this paper, we tackle the problem of 24 hours-monitoring patient actions in a ward such as "stretching an arm out of the bed", "falling out of the bed", where temporal movements are subtle or significant. In the concerned scenarios, the relations between scene layouts and body kinematics (skeletons) become important cues to recognize actions; however they are hard to be secured at a testing stage. To address this problem, we propose a kinematic-layout-aware random forest which takes into account the kinematic-layout (\ie layout and skeletons), to maximize the discriminative power of depth image appearance. We integrate the kinematic-layout in the split criteria of random forests to guide the learning process by 1) determining the switch to either the depth appearance or the kinematic-layout information, and 2) implicitly closing the gap between two distributions obtained by the kinematic-layout and the appearance, when the kinematic-layout appears useful. The kinematic-layout information is not required for the test data, thus called "privileged information prior". The proposed method has also been testified in cross-view settings, by the use of view-invariant features and enforcing the consistency among synthetic-view data. Experimental evaluations on our new dataset PATIENT, CAD-60 and UWA3D (multiview) demonstrate that our method outperforms various state-of-the-arts.
In this paper, we propose a novel single image action recognition algorithm which is based on the idea of semantic body part actions. Unlike existing bottom up methods, we argue that the human action is a combination of meaningful body part actions. In detail, we divide human body into five parts: head, torso, arms, hands and legs. And for each of the body parts, we define several semantic body part actions, e.g., hand holding, hand waving. These semantic body part actions are strongly related to the body actions, e.g., writing, and jogging. Based on the idea, we propose a deep neural network based system: first, body parts are localized by a Semi-FCN network. Second, for each body parts, a Part Action Res-Net is used to predict semantic body part actions. And finally, we use SVM to fuse the body part actions and predict the entire body action. Experiments on two dataset: PASCAL VOC 2012 and Stanford-40 report mAP improvement from the state-of-the-art by 3.8% and 2.6% respectively.
Motivation: Recognizing human actions in a video is a challenging task which has applications in various fields. Previous works in this area have either used images from a 2D or 3D camera. Few have used the idea that human actions can be easily identified by the movement of the joints in the 3D space and instead used a Recurrent Neural Network (RNN) for modeling. Convolutional neural networks (CNN) have the ability to recognise even the complex patterns in data which makes it suitable for detecting human actions. Thus, we modeled a CNN which can predict the human activity using the joint data. Furthermore, using the joint data representation has the benefit of lower dimensionality than image or video representations. This makes our model simpler and faster than the RNN models. In this study, we have developed a six layer convolutional network, which reduces each input feature vector of the form 15x1961x4 to an one dimensional binary vector which gives us the predicted activity. Results: Our model is able to recognise an activity correctly upto 87% accuracy. Joint data is taken from the Cornell Activity Datasets which have day to day activities like talking, relaxing, eating, cooking etc.
Robust visual tracking is a challenging computer vision problem, with many real-world applications. Most existing approaches employ hand-crafted appearance features, such as HOG or Color Names. Recently, deep RGB features extracted from convolutional neural networks have been successfully applied for tracking. Despite their success, these features only capture appearance information. On the other hand, motion cues provide discriminative and complementary information that can improve tracking performance. Contrary to visual tracking, deep motion features have been successfully applied for action recognition and video classification tasks. Typically, the motion features are learned by training a CNN on optical flow images extracted from large amounts of labeled videos. This paper presents an investigation of the impact of deep motion features in a tracking-by-detection framework. We further show that hand-crafted, deep RGB, and deep motion features contain complementary information. To the best of our knowledge, we are the first to propose fusing appearance information with deep motion features for visual tracking. Comprehensive experiments clearly suggest that our fusion approach with deep motion features outperforms standard methods relying on appearance information alone.
Representation of human actions as a sequence of human body movements or action attributes enables the development of models for human activity recognition and summarization. We present an extension of the low-rank representation (LRR) model, termed the clustering-aware structure-constrained low-rank representation (CS-LRR) model, for unsupervised learning of human action attributes from video data. Our model is based on the union-of-subspaces (UoS) framework, and integrates spectral clustering into the LRR optimization problem for better subspace clustering results. We lay out an efficient linear alternating direction method to solve the CS-LRR optimization problem. We also introduce a hierarchical subspace clustering approach, termed hierarchical CS-LRR, to learn the attributes without the need for a priori specification of their number. By visualizing and labeling these action attributes, the hierarchical model can be used to semantically summarize long video sequences of human actions at multiple resolutions. A human action or activity can also be uniquely represented as a sequence of transitions from one action attribute to another, which can then be used for human action recognition. We demonstrate the effectiveness of the proposed model for semantic summarization and action recognition through comprehensive experiments on five real-world human action datasets.
Single modality action recognition on RGB or depth sequences has been extensively explored recently. It is generally accepted that each of these two modalities has different strengths and limitations for the task of action recognition. Therefore, analysis of the RGB+D videos can help us to better study the complementary properties of these two types of modalities and achieve higher levels of performance. In this paper, we propose a new deep autoencoder based shared-specific feature factorization network to separate input multimodal signals into a hierarchy of components. Further, based on the structure of the features, a structured sparsity learning machine is proposed which utilizes mixed norms to apply regularization within components and group selection between them for better classification performance. Our experimental results show the effectiveness of our cross-modality feature analysis framework by achieving state-of-the-art accuracy for action classification on five challenging benchmark datasets.
Convolutional Neural Networks (ConvNets) have recently shown promising performance in many computer vision tasks, especially image-based recognition. How to effectively apply ConvNets to sequence-based data is still an open problem. This paper proposes an effective yet simple method to represent spatio-temporal information carried in $3D$ skeleton sequences into three $2D$ images by encoding the joint trajectories and their dynamics into color distribution in the images, referred to as Joint Trajectory Maps (JTM), and adopts ConvNets to learn the discriminative features for human action recognition. Such an image-based representation enables us to fine-tune existing ConvNets models for the classification of skeleton sequences without training the networks afresh. The three JTMs are generated in three orthogonal planes and provide complimentary information to each other. The final recognition is further improved through multiply score fusion of the three JTMs. The proposed method was evaluated on four public benchmark datasets, the large NTU RGB+D Dataset, MSRC-12 Kinect Gesture Dataset (MSRC-12), G3D Dataset and UTD Multimodal Human Action Dataset (UTD-MHAD) and achieved the state-of-the-art results.
We propose a new deep network for audio event recognition, called AENet. In contrast to speech, sounds coming from audio events may be produced by a wide variety of sources. Furthermore, distinguishing them often requires analyzing an extended time period due to the lack of clear sub-word units that are present in speech. In order to incorporate this long-time frequency structure of audio events, we introduce a convolutional neural network (CNN) operating on a large temporal input. In contrast to previous works this allows us to train an audio event detection system end-to-end. The combination of our network architecture and a novel data augmentation outperforms previous methods for audio event detection by 16%. Furthermore, we perform transfer learning and show that our model learnt generic audio features, similar to the way CNNs learn generic features on vision tasks. In video analysis, combining visual features and traditional audio features such as MFCC typically only leads to marginal improvements. Instead, combining visual features with our AENet features, which can be computed efficiently on a GPU, leads to significant performance improvements on action recognition and video highlight detection. In video highlight detection, our audio features improve the performance by more than 8% over visual features alone.
In recent years, there has been renewed interest in developing methods for skeleton-based human action recognition. A skeleton sequence can be naturally represented as a high-order tensor time series. In this paper, we model and analyze tensor time series with Linear Dynamical System (LDS) which is the most common for encoding spatio-temporal time-series data in various disciplines dut to its relative simplicity and efficiency. However, the traditional LDS treats the latent and observation state at each frame of video as a column vector. Such a vector representation fails to take into account the curse of dimensionality as well as valuable structural information with human action. Considering this fact, we propose generalized Linear Dynamical System (gLDS) for modeling tensor observation in the time series and employ Tucker decomposition to estimate the LDS parameters as action descriptors. Therefore, an action can be represented as a subspace corresponding to a point on a Grassmann manifold. Then we perform classification using dictionary learning and sparse coding over Grassmann manifold. Experiments on MSR Action3D Dataset, UCF Kinect Dataset and Northwestern-UCLA Multiview Action3D Dataset demonstrate that our proposed method achieves superior performance to the state-of-the-art algorithms.
Most successful deep learning algorithms for action recognition extend models designed for image-based tasks such as object recognition to video. Such extensions are typically trained for actions on single video frames or very short clips, and then their predictions from sliding-windows over the video sequence are pooled for recognizing the action at the sequence level. Usually this pooling step uses the first-order statistics of frame-level action predictions. In this paper, we explore the advantages of using higher-order correlations; specifically, we introduce Higher-order Kernel (HOK) descriptors generated from the late fusion of CNN classifier scores from all the frames in a sequence. To generate these descriptors, we use the idea of kernel linearization. Specifically, a similarity kernel matrix, which captures the temporal evolution of deep classifier scores, is first linearized into kernel feature maps. The HOK descriptors are then generated from the higher-order co-occurrences of these feature maps, and are then used as input to a video-level classifier. We provide experiments on two fine-grained action recognition datasets and show that our scheme leads to state-of-the-art results.
Deep learning models have achieved state-of-the- art performance in recognizing human activities, but often rely on utilizing background cues present in typical computer vision datasets that predominantly have a stationary camera. If these models are to be employed by autonomous robots in real world environments, they must be adapted to perform independently of background cues and camera motion effects. To address these challenges, we propose a new method that firstly generates generic action region proposals with good potential to locate one human action in unconstrained videos regardless of camera motion and then uses action proposals to extract and classify effective shape and motion features by a ConvNet framework. In a range of experiments, we demonstrate that by actively proposing action regions during both training and testing, state-of-the-art or better performance is achieved on benchmarks. We show the outperformance of our approach compared to the state-of-the-art in two new datasets; one emphasizes on irrelevant background, the other highlights the camera motion. We also validate our action recognition method in an abnormal behavior detection scenario to improve workplace safety. The results verify a higher success rate for our method due to the ability of our system to recognize human actions regardless of environment and camera motion.
We investigate the problem of representing an entire video using CNN features for human action recognition. Currently, limited by GPU memory, we have not been able to feed a whole video into CNN/RNNs for end-to-end learning. A common practice is to use sampled frames as inputs and video labels as supervision. One major problem of this popular approach is that the local samples may not contain the information indicated by global labels. To deal with this problem, we propose to treat the deep networks trained on local inputs as local feature extractors. After extracting local features, we aggregate them into global features and train another mapping function on the same training data to map the global features into global labels. We study a set of problems regarding this new type of local features such as how to aggregate them into global features. Experimental results on HMDB51 and UCF101 datasets show that, for these new local features, a simple maximum pooling on the sparsely sampled features lead to significant performance improvement.
Understanding human actions in visual data is tied to advances in complementary research areas including object recognition, human dynamics, domain adaptation and semantic segmentation. Over the last decade, human action analysis evolved from earlier schemes that are often limited to controlled environments to nowadays advanced solutions that can learn from millions of videos and apply to almost all daily activities. Given the broad range of applications from video surveillance to human-computer interaction, scientific milestones in action recognition are achieved more rapidly, eventually leading to the demise of what used to be good in a short time. This motivated us to provide a comprehensive review of the notable steps taken towards recognizing human actions. To this end, we start our discussion with the pioneering methods that use handcrafted representations, and then, navigate into the realm of deep learning based approaches. We aim to remain objective throughout this survey, touching upon encouraging improvements as well as inevitable fallbacks, in the hope of raising fresh questions and motivating new research directions for the reader.