search

haimengzhao / CAE-ADMM

CAE-ADMM: Implicit Bitrate Optimization via ADMM-Based Pruning in Compressive Autoencoders
https://arxiv.org/abs/1901.07196
MIT License
46 stars 9 forks source link

dataset and setting #1

Closed oDestroyeRo closed 5 years ago

oDestroyeRo commented 5 years ago

I clone this repo and download BSD500 for the dataset when I run "train.py" the result not a color image like this https://cdn1.imggmi.com/uploads/2019/5/30/53e33bf97ada5a4e1f39bc5f15c2c43b-full.png Can you help me to slove the problem?

haimengzhao commented 5 years ago

Hi. When you train a new model, the outputs are indeed colorless at first. But it will gradually change for the better as the training process goes on. If you are willing to load the pretrained models, you will have to modify the settings in model.py to fit the selected checkpoint and change the path correspondingly in train.py.