Open BoyuanYan opened 6 years ago
We have presented the detailed structure of the encoding and decoding layers in our paper. Please read our paper for more details.
Sorry about reading your paper roughly, and I'll read it in detail later. However, there are still two little questions. I'll be appreciated very much for any reply.
We do not use BN after convolutional layers of the WAE. We merely use one BN layer when feeding the decomposed sub-image into the classification network.
We train the WAE about 20 epoches on the ImageNet dataset.
I could extract the encoding layer on stage 1, but there isn't enough information about decoding layer structure. I've tried the symmetrical structure of encoding layer which contains only one BN after deconv. However, I'm not sure if it is correct. The image reconstruction after 1 epoch shows that high-freq image is likely all-dark, and loss decreases to 131.449 from 10000, which are confusing.