## Prerequisites - Linux or macOS - Python 3 - CPU or NVIDIA GPU + CUDA CuDNN ## Setting enviroments(Installation) -创建环境： conda create --name FA_generation python=3.6 -激活环境： conda activate FA_generation -依赖库安装： pip install -r requirements.txt ### 1. Train/Test (Only CycleGAN) - Train the model : ``` python train.py --dataroot ./datasets/FFA_detect --name FA_detect_cycle --model cycle_gan ``` - Test the model: ``` python test.py --dataroot ./datasets/FFA --name FFA --model cycle_gan --preprocess resize --load_size 512 ``` ### 2. Train/Test (CycleGAN + CBAM) - Train the model : ``` python train.py --dataroot ./datasets/FFA_detect --name FA_detect_CBAM --model cycle_gan --netG attention_resnet_9blocks ``` - Test the model: ``` python test.py --dataroot ./datasets/FFA_detect --name FA_detect_CBAM --model cycle_gan --netG attention_resnet_9blocks ``` ### 3. Train/Test (CycleGAN + CBAM + cam_mask loss) (Final Model!!!) - Train the model : ``` python train.py --dataroot ./datasets/FFA_detect --name FA_detect_CAM --model cycle_gan_cam --netG attention_resnet_9blocks --lambda_mask 10.0 ``` - Test the model: ``` python test.py --dataroot ./datasets/FFA_detect --name FA_detect_CAM --model cycle_gan_cam --netG attention_resnet_9blocks --lambda_mask 10.0 ```