Hi! how to get ground-true labels for another dataset?

switchablenorms / CelebAMask-HQ

A large-scale face dataset for face parsing, recognition, generation and editing.

2.12k stars 352 forks source link

Hi! how to get ground-true labels for another dataset? #15

Closed wqz960 closed 4 years ago

wqz960 commented 5 years ago

Hi! Thank you for your dataset and codes! But if I have a face dataset such as FDDB or 300W, can I get the ground-true labels for these datasets and train their network? Thank you!

steven413d commented 5 years ago

Thanks for your interest in our project! We have no ground-true labels for these datasets. You can try to inference these images on our well-trained parsing model.

wqz960 commented 5 years ago

@steven413d yes! I have tried that way, but the performance is not so good, some images shows very bad result. how did the HQ dataset come？thank you！

steven413d commented 5 years ago

Hi, CelebAMask-HQ was labeled according to CelebA-HQ dataset. Our released parsing model was a simple Unet architecture. You can try a stronger one.