Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 53497 (05)
- University: Sharif University of Technology
- Department: Materials Science and Engineering
- Advisor(s): Ghazizadeh, Ali
- Abstract:
- Looking for eye paths is widely used in various research and even commercial areas, Eye trackers that are used commercially today do this by using infrared transmitters and receivers. As the speed and performance of the processors advanced, many efforts have been made to create an eye-tracking device using visible light without any movement restrictions for the subject, and efforts to increase the accuracy and speed of sampling are still ongoing; The initial methods proposed in this area are feature-based, but newer papers and researches have used Deep learning methods to do this. The commonly used methods for eye tracking in visible light are three main steps: 1. Fetching frames from the high-speed camera 2. Face Detection and Eye Determination 3. Determine the location of the pupil, and our goal in this section is to increase the accuracy and speed of the processing of these algorithms. As expected, these three actions are time-consuming, for example, achieving a device with a frequency of 30 Hz is not so difficult, but when it is necessary to do sampling at a frequency of 2000 Hz, we need to do a few major things: 1. Increase the speed of processing algorithms 2. Hardware Considerations: For this, we have two suggestions, one for hardware implementation, and for example on FPGA boards, or using multithreading on the GPU.
- Keywords:
- Deep Learning ; Image Processing ; Eye Tracking ; Field Programmable Gate Array (FPGA) ; Multithread Systems ; Non-Infrared Eye Tracker
Digital Object List
-
محتواي کتاب
- view
Bookmark