Rare traffic sign detection with synthetic images and multiple classifiers
Detecting rare traffic signs is important for various applications such as autonomous driving, creation of city maps, and road maintenance, as they can provide useful information regarding the surroundings to aid driving decision-making. In this project, we demonstrate that we can train neural netwo...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2020
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/136942 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | Detecting rare traffic signs is important for various applications such as autonomous driving, creation of city maps, and road maintenance, as they can provide useful information regarding the surroundings to aid driving decision-making. In this project, we demonstrate that we can train neural networks for the rare traffic sign detection and classification tasks by making use of synthetically rendered images and data augmentation to overcome the lack of rare traffic sign training samples. We also show that two-stage detectors are advantageous in achieving high recall rates on rare traffic signs over single-stage detectors, and is a promising avenue of research with regards to addressing the rare sign detection problem. |
|---|