The main body of this project comes from v7.0 - YOLOv5 SOTA Realtime Instance Segmentation of ultralytics/yolov5
This version uses GhostNet as the backbone network, and adds CBAM(Coordinate attention) attention mechanism in v3.0.0.
Important Updates
- Replacing the backbone network: CSPDarknet53 -> GhostNet, to make it lightweight🎉NEW
- Add CBAM attention mechanisms🎉NEW
- After three Concats near the Head, 17/21/25Layer.
- Using VOC2007 dataset for training.
Specific Changes
- Add "models/parts_yolov5s-ghostnet_cbam.yaml"🎉NEW
- After three Concats near the Head, 17/21/25Layer in "models/parts_yolov5s-ghostnet_cbam.yaml".
- Modify "train.py"
- line 441: initial weights path -> weights/yolov5s.pt
- line 442: model.yaml path -> models/parts_yolov5s-ghostnet_cbam🎉NEW
- line 443: dataset.yaml path -> data/parts_voc2007.yaml
Notes
The modification of ECA and CA attention mechanisms is the same as CBAM.