https://www.youtube.com/watch?v=y6UmV8QwO9Q&list=PLkRkKTC6HZMy8smJGhhZ4HBIQgShLaTo8
1. Faster RCNN : RPN is trained separately from classification network.
2. Feature extraction layers like VGG are required for RPN, classification. It's applied for both faster rcnn and ssd.
'Papers > Machine learning' 카테고리의 다른 글
| Anchor size for yolo (0) | 2019.04.27 |
|---|---|
| Loss function for yolo (0) | 2019.04.27 |
| Cross-entropy cost function (0) | 2019.04.27 |
| batch, epoch, iteration (0) | 2019.04.15 |
| K-Means clustering (0) | 2019.04.14 |