Questions about data preprocess of VFHQ

[Time-Lord12th]

Thanks for your great work!
I have some questions about data preprocess of VFHQ.

1. Do you use flame_optim_batch_singleframe.py or flame_optim_batch_multiframe.py for full BFM-to-FLAME transformation process?
2. For face segmenation using Mask2Former, which trained model should I choose?
3. If possible, can you tell me the time cost of VFHQ preprocessing？

Thanks!

[gangxu822]

+1 same question

[gangxu822]

https://github.com/facebookresearch/Mask2Former/blob/9b0651c6c1d5b3af2e6da0589b719c514ec0d69a/MODEL_ZOO.md

i use this conf and model,match is import

[gangxu822]

[gangxu822]

python demo.py --config-file ../configs/coco/panoptic-segmentation/maskformer2_R50_bs16_50ep.yaml \
--input "$input_path"
--output "$output_path"
--opts MODEL.WEIGHTS ./model_final_94dc52.pkl