Semantic Segmentation Based Detection and Processing Speed Optimization on an Object Detection Hardware for Autonomous Vehicle Navigation
Autonomous Driving is being proposed as a solution for a multitude of problems related to transportation. It is a necessity for Autonomous Driving platforms to have computing hardware that is able to detect traffic conditions for navigation purposes. We implemented an algorithm based on a fully conv...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/36048 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Autonomous Driving is being proposed as a solution for a multitude of problems related to transportation. It is a necessity for Autonomous Driving platforms to have computing hardware that is able to detect traffic conditions for navigation purposes. We implemented an algorithm based on a fully convolutional model to provide a semantic segmentation of the road scene on embedded hardware. Afterwards, further processing of semantic segmentation output is done to provide warning systems to satisfy the level 0 autonomous driving criteria. This system is then optimized for detection speed. Optimization is done through various graph operations such as folding and other manual transformations. Optimization successfully provides a rise in detection speed performance, and warning systems successfully detects objects in front of the vehicle to provide collisions warnings. |
|---|