Mobile platform control using machine learning and human inputs
Using human input to control a mobile platform has gained popularity in recent years. With the rapid development of machine learning, more and more types of human input can be analyzed by computers to obtain human intention. In this dissertation, human eye images selected from the MGIIGaze dataset a...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis-Master by Coursework |
| Language: | English |
| Published: |
Nanyang Technological University
2023
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/163994 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | Using human input to control a mobile platform has gained popularity in recent years. With the rapid development of machine learning, more and more types of human input can be analyzed by computers to obtain human intention. In this dissertation, human eye images selected from the MGIIGaze dataset are extracted and labeled into five classes according to the gaze direction, corresponding to five commands that a mobile platform can recognize. There are five machine learning models (k-Nearest Neighbors, Random Forest, two types of Convolutional Neural Networks, and Vision Transformer) explored and compared in this dissertation to find the most suitable one for the eye images. K-fold validation and grid research methods are applied to optimize each model. The conclusion is that the Vision Transformer outperforms other models for this dataset. More studies using algorithms combined with Vision Transformer can also be explored in the future with a larger dataset. |
|---|