yaw, pitch, and roll values among the data labels. How are these values determined?

[jeungmin717]

Thank you for organizing the good data. I have a question.

I am curious about the yaw, pitch, and roll values among the data labels. How are these values determined? I understand that for the IMDB dataset, it is not possible to measure the yaw, pitch, and roll values of characters in movies using any instruments or devices. So, I wonder what models or formulas are used to derive these values.

Additionally, I would like to know if the IMDB head pose label data has sufficient accuracy to train a Head Pose Estimation model.