Questions about the training strategy

Hi:

I use the provided template repositories to train the model in detection track, but the model weight trained by myself is much worse than the provided (model.pth), what i only did was oversampled the positive cases to reach the number of negative cases., and i have checked that positive cases are correctly oversampled, are there any other problems may exist？Thank you!

lic

Hi Lic,

We have explained the baseline training strategy in this post.

Best, Ecem