Day34 - NWD在FLIR上的表现

yolov5n/s/m和yolov8n/s仅使用NWD损失函数,在FLIR上训练200epochs

• Yolov5n-nwd

  Parameters: 2.655M

  Class	Precision	Recall	mAP50	mAP50-95
  all	0.599	0.464	0.491	0.286
  person	0.804	0.629	0.735	0.402
  bike	0.467	0.447	0.423	0.252
  car	0.832	0.748	0.825	0.574
  motor	0.799	0.578	0.658	0.333
  bus	0.817	0.45	0.607	0.406
  truck	0.273	0.413	0.164	0.0902
  light	0.72	0.418	0.491	0.219
  other vehicle	0.0771	0.0317	0.025	0.00881

• Yolov5s-nwd

  Parameters: 9.153M

  Class	Precision	Recall	mAP50	mAP50-95
  all	0.651	0.507	0.555	0.338
  person	0.86	0.651	0.781	0.453
  bike	0.556	0.518	0.502	0.296
  car	0.876	0.778	0.863	0.622
  motor	0.745	0.582	0.701	0.384
  bus	0.794	0.536	0.648	0.447
  truck	0.378	0.435	0.294	0.193
  light	0.802	0.458	0.566	0.268
  other vehicle	0.198	0.0952	0.0824	0.0423

• Yolov5m-nwd

  Parameters: 25.111M

  Class	Precision	Recall	mAP50	mAP50-95
  all	0.681	0.517	0.579	0.364
  person	0.884	0.657	0.801	0.486
  bike	0.626	0.541	0.554	0.356
  car	0.906	0.786	0.874	0.642
  motor	0.797	0.636	0.736	0.386
  bus	0.814	0.503	0.69	0.501
  truck	0.377	0.457	0.3	0.213
  light	0.852	0.494	0.623	0.298
  other vehicle	0.19	0.0635	0.0583	0.0295

• Yolov8n-nwd

  Parameters: 3.157M

  Class	Precision	Recall	mAP50	mAP50-95
  all	0.612	0.428	0.49	0.292
  person	0.848	0.613	0.742	0.411
  bike	0.563	0.447	0.44	0.262
  car	0.864	0.738	0.833	0.584
  motor	0.758	0.545	0.643	0.33
  bus	0.752	0.458	0.612	0.431
  truck	0.226	0.216	0.135	0.0924
  light	0.802	0.374	0.495	0.219
  other vehicle	0.0843	0.0317	0.0241	0.0103

• Yolov8s-nwd

  Parameters: 11.167M

  Class	Precision	Recall	mAP50	mAP50-95
  all	0.633	0.501	0.543	0.335
  person	0.849	0.669	0.784	0.457
  bike	0.567	0.541	0.514	0.312
  car	0.878	0.779	0.863	0.621
  motor	0.748	0.6	0.671	0.373
  bus	0.771	0.559	0.66	0.47
  truck	0.269	0.304	0.208	0.134
  light	0.8	0.474	0.58	0.27
  other vehicle	0.181	0.0794	0.0673	0.0416