Yu Sun

看了作者的知乎，官方的yolov5是支持单类别训练的，也就是nc=1的时候是可以正常训练的，看了大佬的loss源码，个人感觉没有问题，特别是类别损失这里，和正框是一致的，也就是说，nc=1的时候是不计算分类损失的. 但是很好奇为什么单类别训练效果非常差.

> if self.nc > 1: # cls loss (only if multiple classes) # t = torch.full_like(ps[:, 5:], self.cn, device=device) # targets t = torch.full_like(ps[:, 5:class_index], self.cn, device=device) # targets t[range(n),...

> @1343744768 so do you know why training nc=1, result is bad? Actually, I don't know. I have checked the source code of the author, it is really unable to...

I am also confused about the model’s output. From the example, it can be seen that if the input is a sequence of 6 frames of images with the shape...