An efficient encoder–decoder model for portrait depth estimation from single images trained on pixel-accurate synthetic data
