insect-detect-v1.4
For the v1.4 release, new YOLOv5n detection models were added. With the YOLOv5n models performance increases significantly to 32 fps for insect detection and 30 fps for detection + tracking on the OAK devices. To reproduce the correct speed (fps) measurement while connected via SSH (X11 forwarding of the frames), print fps to the console and comment out cv2.imshow(), as this will significantly slow down the received message output and thereby fps. If you are using e.g. a Raspberry Pi 4 B connected to a screen, fps will be correctly shown in the livestream.
| Model (.blob) |
size (pixels) |
mAPval 50-95 |
mAPval 50 |
Precisionval |
Recallval |
Speed OAK (fps) |
|---|---|---|---|---|---|---|
| YOLOv5n (5 shaves) | 416 | 58.2 | 97.4 | 97.0 | 95.0 | ~32 |
| YOLOv5n (4 shaves) + tracker | 416 | 58.2 | 97.4 | 97.0 | 95.0 | ~30 |
| YOLOv5s (5 shaves) | 416 | 63.4 | 97.8 | 96.6 | 95.6 | ~17 |
| YOLOv5s (4 shaves) + tracker | 416 | 63.4 | 97.8 | 96.6 | 95.6 | ~17 |
New features
Add new YOLOv5n models: 5fb3e48
Important Updates
Changelog
Changes between previous release and this release: v1.3...v1.4