68-->70 pretrained model for landmark detection on synthetics data

[wenzhengzeng]

Thanks for your great work. I notice that the synthetics dataset that your train the face alignment model has 70 landmark but the ResNet50d model only outputs 68 of them, missing the 2 pupil centers. Can you give a pre-trained model that contains 70 landmarks including 2 pupil centers for a more comprehensive downstream usage？Thanks.

[nttstar]

Hello, right now we don't have enough time to re-train a 70 landmarks model. You can simply follow our training guidelines to reproduce the result.