Closed ghost closed 7 years ago
Please try this branch of my caffe: https://github.com/chenxi116/caffe/tree/feature/20160617_cb_softattention Basically it integrates the DeepLab v2 code into a more recent caffe. Using the original DeepLab v2 code released by the DeepLab author should also work.
Could you send me a copy of ResNet101_init.tfmodel
. I use windows, and it is hard for me to compile a customized caffe. Thanks very much!
I might be able to do that later, but not today. I just thought of another trick that might work. Since the interp layers in deeplab do not have weights, maybe you can try edit the deploy.prototxt and remove all interp layers.
Great thanks. I will try that first.
I am using the windows caffe python3.5 vs2015 release version. But it reports
Thanks!