1.If you are doing a multi-dataset conversion task (this example is to convert on two datasets), you need two description files in txt format, like:
5_o_Clock_Shadow Arched_Eyebrows Attractive Bags_Under_Eyes Bald
000001.jpg -1 1 1 -1
2.To perform more domain conversion tasks(three or more), as described in the paper, you can add 'txt_dir', 'features', 'imfile' to the source code.
3.You can add gradient penalties or other GAN models, which can be achieved with easy code changes.
4.Example of use:
python E:\delete\StarGAN_Multi-domain_tensorflow\main.py --phase train --features_A Black_Hair,Blond_Hair,Brown_Hair,Male,Young --features_B Angry,Disgusted,Fearful,Happy,Neutral --txt1_dir ./data/train.txt --txt2_dir ./data/trainano.txt --imfile1 ./data/train/ --imfile2 ./data/trainano/ --save_path ./data/Trainims/ --ckpt_path ./data/ckpt/model.ckpt --logdir ./data/Log/ --test_ckpt ./data/ckpt/ --test_imsdir ./data/Test/ --test_save_path ./data/Testims/ --test_target Angry,Disgusted,Young --lr 0.0001 --batch_size 20 --epoch 20 --save_iters 20 --ld 1 --lr 0.0001 --adv_weight 1 --cls_weight 10 --recy_weight 10
5.discourse address:https://arxiv.org/abs/1711.09020
Rongpeng-Lin/StarGAN-Multi-domain-use-tensorflow
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
About
No description, website, or topics provided.
Resources
License
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published