一种人脸识别方法、处理器、芯片、电子设备和存储介质
By cropping ultra-high resolution images and extracting facial key points, the face detection model is simplified, solving the problems of high computational cost and neglecting key points, and improving detection accuracy and recognition accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- GUANGDONG OPPO MOBILE TELECOMMUNICATIONS CORP LTD
- Filing Date
- 2021-12-16
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies, face detection and recognition models for ultra-high resolution images have complex structures, high computational demands, and ignore key facial information, leading to detection failures and low recognition accuracy.
By cropping the initial image to be identified, the target image to be identified is obtained. Then, a face detection model is used to extract facial key points and features, and the loss value is calculated to obtain the recognition result. This simplifies the model structure and reduces the amount of computation.
It reduces computational load, improves face detection accuracy and recognition precision, lowers the requirements for electronic device video memory, and solves the problem of detection failure caused by insufficient video memory.
Smart Images
Figure CN116266411B_ABST