search

meteorshowers / RCF-pytorch

Richer Convolutional Features for Edge Detection model in pytorch CVPR2017
510 stars 145 forks source link

请问训练自己的数据集为什么LOSS不收敛，结果图都是全黑或者全白。 #43

Closed xanxuso closed 4 years ago

xanxuso commented 4 years ago

May I ask why the LOSS does not converge when training our own data set, the result graphs are all black or all white.

meteorshowers commented 4 years ago

you can change the learning rate to stable the training process. My training is fine on my computer.

xanxuso commented 4 years ago

yes,I have solved it,thank you

AishuaiYao commented 3 years ago

yes,I have solved it,thank you

hello， could you mind discussion this problem with me？？？