你好,我注意到在training中,keypoint_subnet 的输入是480x480,而detection_subnet的是608x608,请问是出于什么考虑才设置不同的分辨率去训练这两个网络呢? 另外,我训练的keypoint_subnet经过48个epoch,val_loss为0.00312(上一个best为35epoch时的0.00313),这结果可以吧?
你好,我注意到在training中,keypoint_subnet 的输入是480x480,而detection_subnet的是608x608,请问是出于什么考虑才设置不同的分辨率去训练这两个网络呢?
另外,我训练的keypoint_subnet经过48个epoch,val_loss为0.00312(上一个best为35epoch时的0.00313),这结果可以吧?