The invention discloses a
skin problem diagnosis method based on
deep learning face partitioning. The method comprises the steps that multiple face images are collected, faces are marked according toseven partitions including a
forehead area AH, a left orbital area AEL, a right orbital area AER, a
nose bridge area AN, a
left cheek area ACL, a
right cheek area ACR and a jaw area AJ, and a face partition marking
data set is formed; a
deep learning instance segmentation model is trained on the human face partition
annotation data set to enable three deviation functions, namely partition classification
cross entropy LCrossEntropy, partition outer frame positioning accuracy function LDetect and partition
pixel classification accuracy LMask, to have minimum values; a face region is segmented byusing the trained instance segmentation model, and that a
skin problem exists in each partition is determined; a
skin problem is diagnosed according to the regional priori knowledge, and a corresponding treatment scheme is given. According to the method, the
skin problem diagnosis for human face partitioning is realized, the applicability is good, and an intelligent
skin problem partitioning classification diagnosis and treatment scheme is provided on the premise of ensuring the real-time performance.