search

AlamiMejjati / Unsupervised-Attention-guided-Image-to-Image-Translation

Unsupervised Attention-Guided Image to Image Translation
MIT License
329 stars 47 forks source link

test photos and No significant change was seen #12

Open zychen2016 opened 5 years ago

zychen2016 commented 5 years ago

First step: download your pre_trained model and unzip it. Then download dataset use bash ./download_datasets.sh apple2orange Next python3 -m create_cyclegan_dataset --image_path_a='./input/apple2orange/testB' --image_path_b='./input/apple2orange/testA' --dataset_name="apple2orange_test" --do_shuffle=0 Then python3 main.py --to_train=0 --log_dir=./output/AGGAN/exp_01 --config_filename=./configs/exp_02_test.json --checkpoint_dir=./apple2orange/#the path to your pre_trained_model Then get some outputs in output/AGGAN/exp_01/20181210-172320_switch30_thres_0.1,a imgs directory and results_99.html。But in results_99.html,INPUT column and GENERATED_IMAGE column is same，there is no significant change？could you help me?

zychen2016 commented 5 years ago

image

AlamiMejjati commented 5 years ago

Hi, Your logdir should be in exp_02 instead of exp_01. Also be careful not to perform orange to orange or apple to apple mapping which would give similar results to what you have now.