High-resolution artistic face mark point detection method based on deep learning

The invention discloses a high-resolution artistic face mark point detection method based on deep learning. The method comprises the following steps: firstly, preparing an artistic face data set for model training and evaluation and corresponding mark point annotations; respectively obtaining a global low-resolution prediction mark point detection map and a high-resolution prediction mark point detection map of each region; optimizing global low-resolution and regional high-resolution prediction mark point detection images through mean square error loss; and finally, restoring the high-resolution prediction mark point detection image of each region to the global coordinate system to realize face mark point detection. The method shows an outstanding effect in face mark point detection of artworks such as paintings and print paintings. The invention provides a mode of combining a global coder-decoder network and a regional coder-decoder network to realize rough and refined marking of artistic face marking point