Real-Time Action Recognition Using Multi-level Action Descriptor and DNN

Jin, Chan; Dung, Trung; Liu, Mingjie; Kim, Hakil

doi:10.5772/intechopen.76086

Cited by 2 publications

(2 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The average testing duration for each action sample is 21ms for TCG. For real-time action recognition, it is required that the time for processing a new sequence must be less than 20-30 milliseconds to facilitate real-time applications (to achieve a rate of 30 sequences per second (sps)) [41]. It means that for traffic gesture recognition, the proposed system can perform action classification for real-…”

Section: Application Of the Proposed Methods In Traffic Control Gesture Recognition For Self-driving Carsmentioning

confidence: 99%

An Efficient Feature Fusion of Graph Convolutional Networks and Its Application for Real-Time Traffic Control Gestures Recognition

Pham

et al. 2021

IEEE Access

View full text Add to dashboard Cite

Recently, skeleton-based gesture and action recognition have emerged thanks to the progress in human pose estimation. Gesture representation using skeletal data is robust since skeletal data are invariant to the individual's appearance. Among different approaches proposed for skeleton-based action/gesture recognition, Graph Convolutional Network (GCN) and its variations have obtained great attention thanks to its ability to capture the graph essence of the skeletal data. In this paper, we aim to design an efficient scheme using relative joints of skeleton sequences adapted in a GCN framework. Both spatial features (i.e., joint positions) and temporal ones (i.e., the velocity of joints) are combined to form the input of Attention-enhanced Adaptive GCN (AAGCN). The proposed framework deals with limitations of the original AAGCN when it works on challenging datasets with incomplete and noisy skeletal data. Extensive experiments are carried out on three datasets CMDFALL, MICA-Action3D, NTU-RGBD. Experimental results show that the proposed method achieves superior performance compared with existing methods. Moreover, to illustrate the application of the proposed method in real-time traffic control gesture recognition for autonomous vehicles, we have evaluated the proposed method on the TCG dataset. The obtained results show that the proposed method offers real-time computation capability and good recognition results. These results suggest a promising solution to deploy a real-time and robust recognition technique for gesture-based traffic control in autonomous vehicles.

show abstract

Section: Application Of the Proposed Methods In Traffic Control Gesture Recognition For Self-driving Carsmentioning

confidence: 99%

An Efficient Feature Fusion of Graph Convolutional Networks and Its Application for Real-Time Traffic Control Gestures Recognition

Pham

et al. 2021

IEEE Access

View full text Add to dashboard Cite

show abstract

“…This method also known as double fusion scheme is based on learning the correlations between appearance and motion features which are combined to discover abnormal activities. In 2017, Cheng-Bin Jin, Trung Dung Do, Mingjie Liu and Hakil Kim implemented a "multilevel action descriptor" [5], which consists of three levels: posture, locomotion, and gesture level; each of which corresponds to a di↵erent group of sub-actions describing a single human action, for example, eating while sitting using appearance based temporal features with multiple CNN. Considering all the mentioned case studies for anomalous action recognition system, the viable approach to this problem is in implementing a anomalous recognition system using a CNN+RNN model.…”

Section: Literature Surveymentioning

confidence: 99%

Smart Surveillance System for Anomaly Recognition

Kunal¹,

Pranit²,

Atharva³

et al. 2022

ITM Web Conf.

View full text Add to dashboard Cite

Situation awareness is the key to security. Surveillance systems are installed in all places where security is very important. Manually observing all the surveillance footage captured is a monotonous and time consuming task. Security can be defined in different terms in different conditions like violence detection, theft identification, detecting harmful activities etc. In crowded public places the term security covers almost all type of unusual events. To eliminate the tedious manual surveillance we have developed a smart surveillance which will detect an anomaly and alert the user and authority without any human interference. It is a very critical issue in a smart surveillance system to instantly detect an anomalous behaviour in video surveillance system. In this project, a unified framework based on deep neural network framework is proposed to detect anomalous activities. This neural network framework consists of (a) an object detection module, (b) an object discriminator and tracking module, (c) an anomalous activity detection module based on recurrent neural network. The system is a web application where user can apply for three different security services namely motion detection, fall detection and anomaly detection which is applicable for monitoring different environment like homes, roads, offices, schools, shops, etc. On detection of anomalous activity the system will notify the user and responsible authority regarding the anomaly through mail with an anomaly detected frame attachment.

show abstract

Real-Time Action Recognition Using Multi-level Action Descriptor and DNN

Cited by 2 publications

References 31 publications

An Efficient Feature Fusion of Graph Convolutional Networks and Its Application for Real-Time Traffic Control Gestures Recognition

An Efficient Feature Fusion of Graph Convolutional Networks and Its Application for Real-Time Traffic Control Gestures Recognition

Smart Surveillance System for Anomaly Recognition

Contact Info

Product

Resources

About