Jack Hu

Actually, this is how to fix the bug, I guess. https://github.com/ultralytics/yolov5/commit/69ff781ca5fa995ce2eeae215a486ab9f7ff566c

Well,I tried the way I mentioned to fix it, and it worked. Meanwhile, it greatly accelerated the training process and improved the performance on classes with relatively little training data,...

I have the same question,too. Since I need to estimate the bbox localization quality during the training.Maybe I should have a test?I don't know.

I've accomplished an experiment, in the beginning, they're not very correlated (R square about 0.5 in my dataset). After some time of training, it goes to 0.8 and higher. This...

I tested the correlation between iou(target and pred) and top-1 distribution. Maybe you can choose other metrics.

@pjh4993

I met the same problem when running tools/train.py. Waiting... TOrch version1.7.1 cuda 10.0 mmcv version 1.3.8

> Hello, I meet this error as the same. Do you have fix it? > > > I met the same problem when running tools/train.py. Waiting... TOrch version1.7.1 cuda 10.0...

> > > > Hello, I meet this error as the same. Do you have fix it? > > > > > I met the same problem when running tools/train.py....

me,too.waiting...