Training of Yolo CNN to detect sprouts types
- Analyzing CropAndWeed dataset
- Converting Fine24 subset to 1024x1024 Yolo format
- Training Yolo CNN to detect sprouts types
- Annotating the video sample using trained detector
- Using ByteTrack objects tracking algorithm with velocity filtering and temporal class smoothing
This project requires Python version from 3.10 to 3.13.
For environment installation use
pip install -r requirements.txt
Overall for 24 classes:
Precision: 0.80
Recall: 0.75
mAP50: 0.81
mAP50-95: 0.60
GPU: NVIDIA GeForce RTX 3060 Ti
Resolution: 1024x1024
Batch size: 1
Latency: ~28ms
FPS: 35
Original video:
Annotated video:
