Day56 - P234+CBAM改进在FLIR上的表现

yolov5s/m和yolov8s,删除P5检测头+添加P2检测头,在backbone的C3/C2f模块后添加一层CBAM,在FLIR上训练200epochs

• Yolov5s-p234-cbam

  Parameters: 8.626M

  Class	Precision	Recall	mAP50	mAP50-95
  all	0.616	0.538	0.571	0.36
  person	0.856	0.725	0.84	0.511
  bike	0.573	0.529	0.51	0.326
  car	0.864	0.819	0.891	0.657
  motor	0.646	0.632	0.702	0.386
  bus	0.738	0.504	0.657	0.464
  truck	0.292	0.435	0.233	0.164
  light	0.787	0.564	0.657	0.329
  other vehicle	0.17	0.0952	0.0788	0.042

• Yolov5m-p234-cbam

  Parameters: 23.546M

• Yolov8s-p234-cbam

  Parameters: 10.564M

  Class	Precision	Recall	mAP50	mAP50-95
  all	0.656	0.525	0.577	0.363
  person	0.856	0.721	0.839	0.511
  bike	0.639	0.553	0.552	0.362
  car	0.876	0.813	0.893	0.659
  motor	0.708	0.582	0.644	0.347
  bus	0.837	0.508	0.695	0.488
  truck	0.377	0.413	0.266	0.19
  light	0.782	0.561	0.668	0.328
  other vehicle	0.172	0.0462	0.056	0.0174