hello, I am trying to predict the depth directly from a face image, however the final depth is always predicted for different camera pose. How can I retain the camera pose and face angle, that I have in my initial input image? Thank you