search

FVL2020 / ELSFace

The source code of TMM paper "An Efficient Latent Style Guided Transformer-CNN Framework for Face Super-Resolution"
10 stars 3 forks source link

main_test.py problem #4

Open pchoksy opened 1 month ago

pchoksy commented 1 month ago

Hello Author, Thank you for your contribution in a field of face super resolution. I have a doubt that, how will ' from model_full import High2Low_simple ' imported? This function has not defined.

viraj-s15 commented 4 weeks ago

Any update on this? Facing the same issue

9Zage666 commented 3 weeks ago

Any update on this? Facing the same issue

Hey thx for your interest and concerns! Actually, the model_full.py is not defined and used in this project, which also has no correlations with this paper. In this study, we utilize the celebrate and ffhq for metrics measurements with get_int_loader defined in data/dataloader.py, while the real-world performance is shown by real-world face re construction with visual effects. If you want to simply run this project, just delete the from model_full important xxx code line is ok.Meanwhile, low_data is also not defined in data_1.py, which also has no correlations with this paper. If you want to further explore the real-world degradation inferenced by model, you can define a new .py model for try!

viraj-s15 commented 3 weeks ago

Gotcha, Thank you!