用自己的数据集训练模型

Question

修改了一下yaml文件，再次运行时报错
parser.add_argument('--data', type=str, default=ROOT / 'data/abc.yaml', help='dataset.yaml path')

yaml文件内容

train: ./paper_data/train.txt  # train images (relative to 'path') 80 images
val: ./paper_data/val.txt  # train images (relative to 'path') 10 images
nc: 4  # number of classes
names: ['ruins', 'person', 'danger_house', 'car'] # 改成自己的类别

Traceback (most recent call last):
  File "D:/yolov5-master/train.py", line 642, in <module>
    main(opt)
  File "D:/yolov5-master/train.py", line 537, in main
    train(opt.hyp, opt, device, callbacks)
  File "D:/yolov5-master/train.py", line 216, in train
    assert mlc < nc, f'Label class {mlc} exceeds nc={nc} in {data}. Possible class labels are 0-{nc - 1}'
AssertionError: Label class 5 exceeds nc=4 in data\ab.yaml. Possible class labels are 0-3
Process finished with exit code 1

Answer 1

这个明显是train.py代码里变量mlc配置的是5，所以即使yaml里配置的是4，但是依然报错，需要修改train.py里的mlc才可以适配