The scaling factor between the coordinate system of the vertices in obj and the real world coordinate system
Closed this issue · 1 comments
ForrestPi commented
unclearness commented
I used orthogonal projection with pixel-unit in my demo.py because I don't know the camera parameters of a test image taken from the internet. It works fine for a single view.
Although the original paper is designed for a single view, a multi-view fitting would be possible.
If you want to fit the eyelid model to multi-view images, you need to implement proper projection with pre-calibrated camera parameters (intrinsic and extrinsic) of each view.