Parallel Archithcture Implementation for Heterogeneous Deep Learning Network
| Autor: | Juang, Bo-Ren, 莊博任 |
|---|---|
| Rok vydání: | 2019 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 107 In recent years, deep learning has flourished, and deep learning has a wide range of practical applications in life. The use of technology can split complex images into different categories and categories, and the ability to apply them has become the biggest recent boom. Because of the complexity of the overall structure of the object on the image, the objects on the road may have different viewing angles, climate problems, or the backlight source to reduce the object recognition, or overlap with each other and identify errors. To enhance the detection of pedestrians and pedestrians, we have proposed using Yolov2 and Yolov3 to analyze and train data on the road vehicles and pedestrian images, and to detect multiple video streams at the same time, so that objects can be detected at different angles, which is very suitable. For analysis of images with fairly complex backgrounds, overlapping objects can also be found and classified in the distinction. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|