Hi, @zhaoweicai ,
I would like to ask a question that what value of the ratio of negative to positive samples is used in the training of KITTI dataset. Now I set it to 6 in the training, but the trained model is very bad and there are many false positives and large location errors. Do you have any suggestions about the training?
The +/- sample ratio is controlled by bg_multiple in proposal subnet and fg_fraction in detection subnet. Please check the .prototxt files and the code.
Hi, @zhaoweicai , I would like to ask a question that what value of the ratio of negative to positive samples is used in the training of KITTI dataset. Now I set it to 6 in the training, but the trained model is very bad and there are many false positives and large location errors. Do you have any suggestions about the training?