search

YanzuoLu / CFLD

[CVPR 2024 Highlight] Coarse-to-Fine Latent Diffusion for Pose-Guided Person Image Synthesis
MIT License
183 stars 12 forks source link

Reproduce question on test split #27

Closed RuijieH closed 4 months ago

RuijieH commented 4 months ago

Hi. When i use the checkpoint offered and run the pose_transfer_test.sh using the default hyperparam, I can't generate the same picture you offered in README, also results in poor performance in metric compared to the images you offered. So i want to know are there some changes of hyperparams when you generate results images compared to the default hyperparam

zcai0612 commented 4 months ago

Sharing the same problem

YanzuoLu commented 4 months ago

Hi, @RuijieH To compare with the non-open-sourced ICCV 2023 paper PoCoLD and achieve higher quantitative metrics, we have earlystop strategy during training and the released checkpoints are simply from the final epoch just for demonstration. You could try retrain for fully reproduce and conduct thorough experiments on intermediate states. Good luck！