Dynamic Sampling Networks for Efficient Action Recognition in Videos

Action recognition in still images is closely related to various other computer vision tasks such as pose estimation, object recognition, image retrieval, video action recognition and frame tagging in videos. This problem is focused on recognizing a person's action or behavior using a single frame. Unlike action recognition in videos - a relatively very well established area of research where spatio-temporal features are used, these are not available for still images, making the problem more challenging. In the present work only actions that involve objects are considered. A complex action is broken down into components based on semantics. The importance of each of these components in action recognition is systematically studied.

Understanding semantic meaning of human actions captured in unconstrained environments has broad applications in fields ranging from patient monitoring, human-computer interaction, to surveillance systems. However, while great progresses have been achieved on automatic human action detection and recognition in videos that are captured in controlled/constrained environments, most existing approaches perform unsatisfactorily on videos with uncontrolled/unconstrained conditions (e.g., significant camera motion, background clutter, scaling, and light conditions). To address this issue, the authors propose a robust human action detection and recognition framework that works effectively on videos taken in controlled or uncontrolled environments. Specifically, the authors integrate the optical flow field and Harris3D corner detector to generate a new spatial-temporal information representation for each video sequence, from which the general Gaussian mixture model (GMM) is learned. All the mean vectors of the Gaussian components in the generated GMM model are concatenated to create the GMM supervector for video action recognition. They build a boosting classifier based on a set of sparse representation classifiers and hamming distance classifiers to improve the accuracy of action recognition. The experimental results on two broadly used public data sets, KTH and UCF YouTube Action, show that the proposed framework outperforms the other state-of-the-art approaches on both action detection and recognition.

Real-time action recognition on smartphones and wearable cameras is a challenging task. The effect of camera motion on action recognition algorithms is not trivial. To our advantage, smartphones and wearable devices are often augmented with depth, audio, geo-location, and inertial data, which can be incorporated into action recognition framework. Our study aims to quantify the effect of ego-centric motion on standard action recognition algorithms. As a part of the study, we collected multi-sensor video dataset with seven actions from the same subjects captured in three different camera motion scenarios—minimal, pure rotational and both rotational and translational. We plan to create a multi-sensor fusion based action recognition framework to improve the recognition accuracy on smartphones and wearable cameras, which are equipped with inertial sensors. We present preliminary experiment results on our multi-sensor video dataset.

Action recognition in videos is currently in the focus of scientific research due to improvements made in automatic analysis of static images and greater availability of processing power. The paper provides an overview of the key models and methods for action recognition that comprise human models and methods based on estimation of joint trajectories, silhouettes and template matching and spatio-temporal local descriptors. To deal with compound actions and activities, action semantic models are proposed with help of expert knowledge. Since the action recognition task is domain dependent, the methods and models are built and tested on domain specific databases. The paper provides an overview and description of recent video datasets that were created for developing action recognition methods, with an emphasis on datasets with additional modalities such as depth images or accelerometer data.

Existing RGB and CNN-based methods in video action recognition mostly do not distinguish human body from the environment, thus easily overfit the scenes and objects of training sets. In this work, we present a conceptually simple, general and high-performance framework for action recognition in videos, aiming at person-centric modeling. The method, called Action Machine, is based on person bounding boxes for instance-level action analysis. It extends the Inflated 3D ConvNet (I3D) by adding a branch for human pose estimation and a 2D CNN for pose-based action recognition. Action Machine can benefit from the multi-task training of action recognition and pose estimation, the fusion of predictions from RGB images and poses. Experiments results are provided on trimmed video action datasets, NTU RGB+D, Northwestern UCLA Multiview Action3D, MSR Daily Activity3D. Action Machine achieves superior performance and generalizes well across datasets.

The task of action recognition in dark videos is useful in various scenarios, e.g., night surveillance and self-driving at night. Though progress has been made in action recognition task for videos in normal illumination, few have studied action recognition in the dark, partly due to the lack of sufficient datasets for such a task. In this paper, we explored the task of action recognition in dark videos. We bridge the gap of the lack of data by collecting a new dataset: the Action Recognition in the Dark (ARID) dataset. It consists of 3,784 video clips with 11 action categories. To the best of our knowledge, it is the first dataset focused on human actions in dark videos. To gain further understanding of our ARID dataset, we analyze our dataset in detail and showed its necessity over synthetic dark videos. Additionally, we benchmark the performance of current action recognition models on our dataset and explored potential methods for increasing their performances. We show that current action recognition models and frame enhancement methods may not be effective solutions for the task of action recognition in dark videos (data available at https://xuyu0010.github.io/arid).

The bag-of-words approach with local spatio-temporal features have become a popular video representation for action recognition in videos. Together these techniques have demonstrated high recognition results for a number of action classes. Recent approaches have typically focused on capturing global statistics of features. However, existing methods ignore relations between features and thus may not be discriminative enough. Therefore, we propose a novel feature representation which captures statistics of pairwise co-occurring local spatio-temporal features. Our representation captures not only global distribution of features but also focuses on geometric and appearance (both visual and motion) relations among the features. Calculating a set of bag-of-words representations with different geometrical arrangement among the features, we keep an important association between appearance and geometric information. Using two benchmark datasets for human action recognition, we demonstrate that our representation enhances the discriminative power of features and improves action recognition performance.

Human action recognition (HAR) has emerged into a prominent area of research in the field of computer vision recently. In this paper, we introduce a novel framework that uses a sparsity-inducing supervised shared dictionary for activity recognition in videos. We demonstrate that sparse coding, also known as sparse dictionary learning, possesses a better discriminating nature that helps to a great extent in classifying action videos into their respective action classes. To achieve robust action recognition, we introduce a novel classification metric, termed Class Selection Index (CSI). We also show that the proposed classification model maintains the accuracy of classification on the Weizmann dataset when compared to other state-of-the-art techniques that have been reported so far, while reducing the computational complexity significantly.

Every year, new Convolutional Neural Network (CNN) architectures appear to deal with different problems in the activity of image and video recognition. These architectures usually work along the ImageNet dataset for looking for the best performance of the CNNs without taking into account the video task where they are used. This can represent a problem if the task is Human Action Recognition (HAR) in video, since the CNN architectures are pre-trained with an image dataset that can practically contain any object, while HAR problem requires consecutive frames of people doing actions. To prove the idea that using CNNs pre-trained on an image dataset does not always achieve the best performance on a video dataset and that, therefore, it is worth comparing the performance of different CNNs under similar circumstances for the HAR problem, this work proposes an analysis between eight different CNN architectures. Each one of the CNN was exclusively trained with RGB images, which were extracted from the frames of the different classes of videos of HMDB51 dataset. To make the classification of an activity in video, we average the predictions taking into account the successes. We also made some ensembles with the best performance CNNs to measure the improvement in accuracy. Our results suggest that Xception is a strong baseline model that could be used by the community to make their comparisons of their proposals more robust.

Activity recognition in videos in a deep-learning setting---or otherwise---uses both static and pre-computed motion components. The method of combining the two components, whilst keeping the burden on the deep network less, still remains uninvestigated. Moreover, it is not clear what the level of contribution of individual components is, and how to control the contribution. In this work, we use a combination of CNN-generated static features and motion features in the form of motion tubes. We propose three schemas for combining static and motion components: based on a variance ratio, principal components, and Cholesky decomposition. The Cholesky decomposition based method allows the control of contributions. The ratio given by variance analysis of static and motion features match well with the experimental optimal ratio used in the Cholesky decomposition based method. The resulting activity recognition system is better or on par with existing state-of-the-art when tested with three popular datasets. The findings also enable us to characterize a dataset with respect to its richness in motion information.

Understanding human behavior and human action recognition are both essential components of effective surveillance video analysis for the purpose of guaranteeing public safety. However, existing approaches such as three-dimensional convolutional neural networks (3D CNN) and two-stream neural networks (2SNN) have computational hurdles due to the significant parameterization they require. In this paper, we offer HARNet, a specialized lightweight residual 3D CNN that is built on directed acyclic graphs and was created expressly to handle these issues and achieve effective human action detection. The suggested method presents an innovative pipeline for creating spatial motion data from raw video inputs, which makes successful latent representation learning of human motions easier to accomplish. This generated input is then supplied into HARNet, which processes spatial and motion information in a single stream in an effective manner, maximizing the benefits of both types of cues. The use of traditional machine learning classifiers is done in order to further improve the discriminative capacity of the features that have been learned. To be more specific, we use the latent representations that are stored in HARNet’s fully connected layer and use them as our deep learnt features. After that, these features are entered into the Support Vector Machine (SVM) classifier in order to accomplish action recognition. In order to evaluate the HARNet-SVM method that was developed, empirical tests were run on commonly used action recognition datasets such as UCF101, HMDB51, and the KTH dataset. These tests were carried out in order to gather data for the evaluation. The experimental results show that our method is superior to other state-of-the-art approaches, achieving considerable performance increases of 2.75% on UCF101, 10.94% on HMDB51, and 0.18% on the KTH dataset. These results were obtained by running the method on each dataset separately. Our findings demonstrate the usefulness of HARNet’s lightweight design and highlight the significance of utilizing SVM classifiers with deep learnt features for the purpose of accurate and computationally efficient human activity recognition in surveillance videos. This work helps to the advancement of surveillance technology, which in turn makes video analysis in applications that take place in the real world safer and more dependable.

Human activity recognition in unconstrained RGB--D videos has extensive applications in surveillance, multimedia data analytics, human-computer interaction, etc, but remains a challenging problem due to the background clutter, camera motion, viewpoint changes, etc. We develop a novel RGB--D activity recognition approach that leverages the dense trajectory feature in RGB videos. By mapping the 2D positions of the dense trajectories from RGB video to the corresponding positions in the depth video, we can recover the 3D trajectory of the tracked interest points, which captures important motion information along the depth direction. To characterize the 3D trajectories, we apply motion boundary histogram (MBH) to depth direction and propose 3D trajectory shape descriptors. Our proposed 3D trajectory feature is a good complementary to dense trajectory feature extracted from RGB video only. The performance evaluation on a challenging unconstrained RGB--D activity recognition dataset, i.e., Hollywood 3D, shows that our proposed method outperforms the baseline methods (STIP-based) significantly, and achieves the state-of-the-art performance.

The recognition of human activities using vision-based techniques has become a crucial research field in video analytics. Over the last decade, there have been numerous advancements in deep learning algorithms aimed at accurately detecting complex human actions in video streams. While these algorithms have demonstrated impressive performance in activity recognition, they often exhibit a bias towards either model performance or computational efficiency. This biased trade-off between robustness and efficiency poses challenges when addressing complex human activity recognition problems. To address this issue, this paper presents a computationally efficient yet robust approach, exploiting saliency-aware spatial and temporal features for human action recognition in videos. To achieve effective representation of human actions, we propose an efficient approach called the dual-attentional Residual 3D Convolutional Neural Network (DA-R3DCNN). Our proposed method utilizes a unified channel-spatial attention mechanism, allowing it to efficiently extract significant human-centric features from video frames. By combining dual channel-spatial attention layers with residual 3D convolution layers, the network becomes more discerning in capturing spatial receptive fields containing objects within the feature maps. To assess the effectiveness and robustness of our proposed method, we have conducted extensive experiments on four well-established benchmark datasets for human action recognition. The quantitative results obtained validate the efficiency of our method, showcasing significant improvements in accuracy of up to 11% as compared to state-of-the-art human action recognition methods. Additionally, our evaluation of inference time reveals that the proposed method achieves up to a 74× improvement in frames per second (FPS) compared to existing approaches, thus showing the suitability and effectiveness of the proposed DA-R3DCNN for real-time human activity recognition.

International audience

Convolutional Neural Networks(CNNs) have achieved great success for object recognition in still images. However, CNNs can't make evident improvement for action recognition in videos, one reason is that many current network architectures are relatively shallow compared with deep models in image domain, and the other reason is that CNNs can't capture effective long-term motion information from videos. Encouraged by the good performance of Residual Network-s(ResNets) for training extremely deep models, and Long-term Recurrent Convolutional Networks(LSTMs) for dealing with tasks involving sequences, we presented an action recognition method based on a two-stream architecture, with 2D ResNets with LSTMs in one stream and designed 3D ResNets with LSTMs in the other stream, which can combine appearance and motion information better. Especially, our proposed method first learns spatiotemporal features of videos through the Residual networks, then models complex temporal dynamics by the Long-term Recurrent Convolutional networks, and with a softmax layer on the top of two streams, the final classification results can be predicted by fusing scores of each stream with weights on score distribution. Furthermore, for better reducing the influence of redundant background information in videos for recognition results, we also applied a center extraction method to generate central regions of videos instead of an entire video into a visual representation. On two video action benchmarks of UCF101 and HMDB51, our method achieved promising performance compared with state-of-the-art.