search

xmojiao / deeplab_v2

基于v2版本的deeplab,使用VGG16模型，在VOC2012，Pascal-context，NYU-v2等多个数据集上进行训练
162 stars 103 forks source link

the result images of mat2png are black, is there some trick? #6

Open luyunxi opened 6 years ago

luyunxi commented 6 years ago

I used my trainning data to fineturning the model. After trainning, the result images are totally black.

xmojiao commented 6 years ago

@luyunxi Maybe you should try to change Makefile.config. Let USE_CUDNN =1 意思就是仍然使用cuda,但是禁止cudnn。