Human body context attention mechanism-based face detection method

The invention discloses a human body context attention mechanism-based human face detection method, relates to the technical field of human face detection, and solves the technical problem of poor precision of an existing detection mode under the view angle of an unmanned aerial vehicle in outdoor construction, and the method comprises the steps: S1, collecting a picture, and obtaining a plurality of regression target frames possibly containing human faces through network PNet regression; S2, filtering target frames with more coincidence in the regression target frames by using soft-NMS to obtain a primary filtering target frame; S3, expanding the primary filtering target frame and processing the primary filtering target frame by using a network Rnet to obtain a high-precision target frame; S4, filtering out the target frames with more coincidence in the high-precision target frames by using soft-NMS to obtain a secondary filtering target frame; S5, expanding the secondary filtering target frame, and processi