I didn't change any lines of codes, but after running your codes and testing it with provided scripts, I will get 100% MR. I am not sure whether you have got State-of-art performance on this datasets? Thank you @CharlesShang During training process, RPN_cls loss did not decrease, it keeps to be around 0.50, seems to be randomly guessing.
I didn't change any lines of codes, but after running your codes and testing it with provided scripts, I will get 100% MR. I am not sure whether you have got State-of-art performance on this datasets? Thank you @CharlesShang During training process, RPN_cls loss did not decrease, it keeps to be around 0.50, seems to be randomly guessing.