How do you use the Cityscape for training?

fyu / dilation

Dilated Convolution for Semantic Image Segmentation

https://www.vis.xyz/pub/dilation

MIT License

781 stars 265 forks source link

How do you use the Cityscape for training? #35

Open MiniBullLab opened 7 years ago

MiniBullLab commented 7 years ago

I use your front-end code to train on the Cityscape dataset, I notice that you add a upsampling layer in dilation10_cityscapes_deploy.prototx, but the front-end_vgg_train_net.prototxt that I have generated don't have this layer, can you tell me the difference?

I am looking forward for your reply, thank you

xingbotao commented 7 years ago

In context,how to compute the height and weight of label_shape?I have a error in train context

09876qwrte commented 7 years ago