Vision based human action recognition using deep learning on an embedded platform | |
| Author | Wickramatilake, Rajapaksha Mudiyanselage Pramod |
| Call Number | AIT Thesis no.CS-21-06 |
| Subject(s) | Deep learning (Machine learning) Human activity recognition Older people--Care--Technological innovations |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Computer Science |
| Publisher | Asian Institute of Technology |
| Abstract | Human action recognition is an exciting field in computer vision. It is beneficial for ac tivity monitoring and human behaviour analysis, especially in the context of elder care. Throughout the last decade, we have seen improvements in video action recognition due to the emergence of deep learning. However, deploying such a system on an embedded plat form is a novel idea that needs to be studied further. This thesis studies the implementation and deployment of a human action recognition system on an embedded device to be used for purposes of elder care. The performance and effectiveness of the system in monitoring crucial actions such as fall-down/unstable movements as well as daily activities such as ex ercising are examined. After several iterations, I obtained an action classification model that got an accuracy of 89% on a challenging test dataset and that runs at 9.0 frames per second on a Nvidia Jetson TX2. |
| Year | 2021 |
| Type | Thesis |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Computer Science (CS) |
| Chairperson(s) | Dailey, Matthew N. |
| Examination Committee(s) | Mongkol Ekpanyapong;Chaklam Silpasuwanchai |
| Scholarship Donor(s) | AIT Partial Scholarship |
| Degree | Thesis (M. Eng.) - Asian Institute of Technology, 2021 |