Authors: Zhou Xinming, Chong Kay Heen, Yoke Kai Xin, Li Zhuoran, Jiang Yuxin, Tan Yuanhong
Email: e0273876@u.nus.edu, e0201414@u.nus.edu, e0148570@u.nus.edu, e0376940@u.nus.edu, e0261921@u.nus.edu, e0261956@u.nus.edu
- Download datasets: sketch2pokemon
- Train the model:
python main.py
- Test the model
- rename the latest checkpoint in
checkpoints
folder intolatest.ckpt
- run
test.ipynb
- rename the latest checkpoint in
Note: training will pick up the latest checkpoint by the name latest.ckpt
- Files for training/testing the model:
- Network definitions for D and G:
discriminators.py
andgenerators.py
- CycleGAN architecture:
model.py
- Helper functions:
ops.py
,utils.py
,extractGreyScale.py
(converting colored trainB to greyscale trainA)
- Network definitions for D and G:
- Files for evaluating the model
eval_utils.py
test.ipynb