YOLOv8-ND: New Detection Algorithm for Fast and Lightweight Object Detection
- Train:
yolov8nd detect train model=yolov8ndn.pt data=coco128.yaml epochs=50
- Validation:
yolov8nd detect val model=yolov8ndn.pt data=coco128.yaml
- Predict:
yolov8nd detect predict model=yolov8ndn.pt source=https://ultralytics.com/images/bus.jpg
- CPU: i5-13500 (E-Core: 2.0GHz, P-Core: 3.0GHz)
- Memory: DDR4 64GB (3200MHz, 128bit width)
- GPU: RTX 4070 (max 2.4GHz, 125W)
- OS: Ubuntu 22.04 LTS
Models | Size | mAP50-95 (%) |
Params (M) |
GFLOPs | Speed GPU(ms) |
Speed CPU(ms) |
---|---|---|---|---|---|---|
YOLOv8ndn | 640 | 36.4 | 3.2 | 7.6 | 4.1 | 36.9 |
YOLOv8nds | 640 | 44.0 | 12.3 | 27.1 | 4.2 | 88.2 |
YOLOv8ndm | 640 | 48.9 | 26.9 | 74.9 | 6.1 | 205.5 |
YOLOv8ndl | 640 | 51.1 | 43.5 | 157.1 | 9.6 | 390.5 |
YOLOv8ndn-lite | 640 | 34.9 | 2.4 | 5.9 | 3.6 | 30.9 |
YOLOv8nds-lite | 640 | 43.2 | 9.2 | 21.1 | 3.7 | 73.7 |
YOLOv8ndm-lite | 640 | 48.5 | 22.2 | 63.7 | 5.3 | 181.4 |
YOLOv8ndl-lite | 640 | 50.4 | 38.2 | 139.7 | 8.6 | 352.4 |
YOLOv8ndn-aux | 640 | 37.5 | 3.7 | 8.7 | 4.4 | 40.7 |
YOLOv8nds-aux | 640 | 45.2 | 14.4 | 31.3 | 4.5 | 99.5 |
YOLOv8ndm-aux | 640 | 49.4 | 33.6 | 91.7 | 7.2 | 245.1 |
YOLOv8ndl-aux | 640 | 51.0 | 56.3 | 198.0 | 11.8 | 485.0 |
YOLOv8n | 640 | 37.5 | 3.2 | 8.9 | 4.4 | 40.9 |
YOLOv8s | 640 | 44.7 | 11.2 | 28.6 | 4.5 | 91.5 |
YOLOv8m | 640 | 50.1 | 25.9 | 78.9 | 6.5 | 215.1 |
YOLOv8l | 640 | 52.9 | 43.7 | 165.2 | 10.1 | 406.5 |
Train parameters: batch=16 epochs=500 optimizer=SGD lr0=0.01 momentum=0.937 data=coco.yaml
Validation parameters: data=coco.yaml batch=1 device=0|cpu
Template from https://github.com/ultralytics/ultralytics (version: 8.2.60)