Dataset: Anime GAN Lite | Kaggle
I manually delete the desktop.ini file
this project is a simple implementation of DC-GAN to generate pictures by random noise.
- configurate the model you want to use in the
config.py
- run
generage.py
orpython generate.py
in the terminal
- put the dataset you want to generate under
./data
folder. - configurate the
config.py
change the argument--dataset
to your own dataset path - run
train.py
orpython train.py
in the terminal
⚠️ Note: since normalization operation included, the color style between generate pic and original pic may not same.
- The output of the discriminator can be optimized, may improve the networks performance
- It still has mode collapse problems in given model. I will fix it later.