wasidennis / AdaptSegNet

Learning to Adapt Structured Output Space for Semantic Segmentation, CVPR 2018 (spotlight)
845 stars 205 forks source link

About the structure of the discriminator #62

Closed jsdd25 closed 4 years ago

jsdd25 commented 5 years ago

@wasidennis Hi, your work is amazing! I have some questions. How you set 5 convolution layers in discriminator? and how you determined it is a proper number of layers? Thank you very much!

hfslyc commented 5 years ago

Hi,

We simply follow the DCGAN's discriminator structure (https://arxiv.org/abs/1511.06434) and make it fully-convolutional.