/dehaze-GAN

Unpaired Training Strategy for Dehazing in Land and Underwater Scenes Using Generative Adversarial Networks

Primary LanguagePython

dehaze-GAN

演算法說明

除霧架構

生成器

Unet1 Unet2

分辨器

training epoachs

使用方法

訓練

python main.py --phase train --pretrain (True or False) --weights_path ./weights_dir/weights.ckpt --dataset_path_x ./dataset_x.npy --dataset_path_y ./dataset_y.npy  --output_dir ./output/

測試

python main.py  --phase test --pretrain True --weights_path ./weights_dir/weights.ckpt --dataset_path_x ./dataset_x.npy --dataset_path_y ./dataset_y.npy  --output_dir ./output/

圖片除霧

python main.py  --phase images --pretrain True --weights_path ./weights_dir/weights.ckpt --imlist ["./sample1.jpg","./sample2.png"] --output_dir ./output/

影片除霧

python main.py  --phase video --pretrain True --weights_path ./weights_dir/weights.ckpt --video_path ./video.avi --output_dir ./output/

結果展示

圖片除霧結果

input output input output

影片除霧結果