The invention relates to a stick figure generation method based on a face photo, belonging to the field of image processing. The method comprises steps of collecting a face database, and using an on-line open source database as a training object; training a face recognition and key point detection model by using the collected database; extracting feature points from a real face picture; sketchingthe input picture; positioning and segmenting the face, five sense organs, hair, clothing and the like based on the extracted feature points; smoothing, enhancing, graying and binarizing the segmentedimage; and recombining the processed components to form a final result. Compared with the prior art, the method has the advantages that the five sense organs are positioned more accurately; a large number of samples are not needed for training, and less picture data is needed; the training time is shorter, the requirement for hardware is low, no cost increases; the training model is very sufficient, and error-prone details are avoided; and the method can be provided to chocolate 3D printers for printing and can be widely used.