Javascript must be enabled for the correct page display

Learning Deep Spatio-Temporal Features for Human Activity Classification

VATTIMUNDA PURAYIL, S (2020) Learning Deep Spatio-Temporal Features for Human Activity Classification. Master's Thesis / Essay, Artificial Intelligence.

[img]
Preview
Text
mAI_2020_VPSubilal.pdf

Download (8MB) | Preview
[img]
Preview
Text
toestemming.pdf

Download (142kB) | Preview

Abstract

The last decades have seen a growing interest and demand for Human Action Recognition (HAR) systems. HAR is widely used, and its presence can be seen in the application areas ranging from Human-Computer interaction (HCI), robotic interactions to video surveillance, autonomous driving, home-based rehabilitation, etc. While offering attractive opportunities for modern systems development, existing HAR systems often fail to recognize human actions precisely. In this thesis, we propose a new model for HAR using deep learning technologies. Our model encompasses the 3D Convolutional Neural Network, which learns the local spatio-temporal features from video input; and a bidirectional Long Short-Term Memory, which learns the long term temporal dependency from the spatio-temporal features to classify the actions. We perform a comparative study of the proposed model with six other baseline models implemented using deep learning methods with different modalities. Extensive experiments demonstrate that our proposed model outperforms baseline models and other state-of-the-art approaches in terms of accuracy. We believe that our work is a step closer to developing accurate real-world HAR applications.

Item Type: Thesis (Master's Thesis / Essay)
Supervisor name: Mohades Kasaei, S.H. and Wiering, M.A.
Degree programme: Artificial Intelligence
Thesis type: Master's Thesis / Essay
Language: English
Date Deposited: 09 Dec 2020 13:07
Last Modified: 09 Dec 2020 13:07
URI: http://fse.studenttheses.ub.rug.nl/id/eprint/23688

Actions (login required)

View Item View Item