About the evaluation metrics of IMavatar

[zliucz]

Hi yufeng,
Your work IMavatar is brilliant 🙂. I re-run your work IMavatar on RTX2080Ti. However, the results of my run and your given pre-trained model seem inconsistent with your paper. Attached is part of my results. I think it is because we use different evaluation metrics codes. My metrics are modified from NerFACE. Would you please share your implementation code of evaluation metrics?

Thanks a lot,
Zichen

[zhengyuf]

Hey Zichen,

Thanks for the kind words!
I've cleaned the metric file and added it as ./code/utils/calculate_metrics.py. I know I've already replied to your email but just replying so that I can close this issue :)

Best,
Yufeng