损失随 YOLOv4-tiny 在 1700 个大数据集上的 16 个类而波动

如何解决损失随 YOLOv4-tiny 在 1700 个大数据集上的 16 个类而波动

我正在使用 YOLOv4-tiny 构建徽标检测系统。我已经构建了一个自定义合成数据集，其中我在游戏屏幕顶部绘制了透明徽标。徽标在绘制到背景上之前进行了增强（模糊、透视变换、调整大小、旋转等）。以下是我拥有的此类数据的一些示例。

我在 1,700 张图片的背景图片上的定义位置粘贴了 1-5 个徽标（随机）。每个标志是一个班级，我有 16 个班级。当我使用 YOLOv4-tiny 运行它时，这是我的输出示例。我的损失在波动，我不明白为什么。

v3 (iou loss,Normalizer: (iou: 0.07,obj: 1.00,cls: 1.00) Region 37 Avg (IOU: 0.281087),count: 1,class_loss = 0.738789,iou_loss = 15.317201,total_loss = 16.055990 
 total_bbox = 485600,rewritten_bbox = 1.175453 % 
v3 (iou loss,cls: 1.00) Region 30 Avg (IOU: 0.547854),count: 5,class_loss = 4.434289,iou_loss = 2.494116,total_loss = 6.928406 
v3 (iou loss,cls: 1.00) Region 37 Avg (IOU: 0.767385),class_loss = 1.036910,iou_loss = 0.562877,total_loss = 1.599788 
 total_bbox = 485606,rewritten_bbox = 1.175439 % 

 (next mAP calculation at 3147 iterations) 
 Last accuracy mAP@0.5 = 3.82 %,best = 3.82 % 
 3038: 2.917786,2.895705 avg loss,0.002610 rate,2.421050 seconds,145824 images,42.306447 hours left
Loaded: 1.968205 seconds - performance bottleneck on CPU or Disk HDD/SSD
v3 (iou loss,cls: 1.00) Region 30 Avg (IOU: 0.602211),count: 3,class_loss = 2.421113,iou_loss = 0.208666,total_loss = 2.629779 
v3 (iou loss,cls: 1.00) Region 37 Avg (IOU: 0.000000),class_loss = 0.000067,iou_loss = 0.000000,total_loss = 0.000067 
 total_bbox = 485609,rewritten_bbox = 1.175431 % 
v3 (iou loss,cls: 1.00) Region 30 Avg (IOU: 0.583863),count: 4,class_loss = 3.290887,iou_loss = 0.876662,total_loss = 4.167549 
v3 (iou loss,cls: 1.00) Region 37 Avg (IOU: 0.383844),class_loss = 0.823941,iou_loss = 8.962539,total_loss = 9.786480 
 total_bbox = 485614,rewritten_bbox = 1.175419 % 
v3 (iou loss,cls: 1.00) Region 30 Avg (IOU: 0.566122),count: 11,class_loss = 9.910147,iou_loss = 1.705420,total_loss = 11.615566 
v3 (iou loss,class_loss = 0.000120,total_loss = 0.000120 
 total_bbox = 485625,rewritten_bbox = 1.175393 % 
v3 (iou loss,cls: 1.00) Region 30 Avg (IOU: 0.582394),class_loss = 3.849459,iou_loss = 0.561072,total_loss = 4.410531 
v3 (iou loss,cls: 1.00) Region 37 Avg (IOU: 0.365837),class_loss = 0.918538,iou_loss = 6.407685,total_loss = 7.326223 

我的问题是，这有什么收获？我什至没有针对测试集测试模型以查看它的性能如何。我需要改进徽标的增强吗？我如何理解这个输出？

更新 这就是模型的表现。它正确地检测到它是什么标志，但在边界框上完全搞砸了。