Face recognition model acquisition method and system for video monitoring scene, and medium

• Main Authors: LI YILI, SHAO XINQING, LIU QIANG, SONG YONGJUN
• Format: Patent
• Language: Chinese; English
• Published: 10.01.2020
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention relates to the technical field of face recognition, in particular to a face recognition model acquisition method and system for a video monitoring scene. The method comprises the following steps: firstly, training a teacher neural network model by adopting a high-resolution image set with a face label; and then obtaining a student network model with the same structure as the teacherneural network model, performing initial training and fine tuning training on the student network model in sequence to obtain an advanced face recognition model, and performing face recognition on theinput video data by adopting the obtained advanced face recognition model. Therefore, the advanced face recognition model obtained through training is obtained based on a high-precision teacher network model, the recognition precision of video information such as small faces, fuzzy faces and side faces in a video monitoring scene is better, and the face information can be accurately recognized. And meanwhile, the advanced