人脸识别模型训练方法、装置、电子设备及存储介质
By combining sample pair learning and prototype learning, and using quality scores to perform weighted updates on the face recognition model, the problems of poor face recognition performance and low accuracy in existing technologies are solved, thereby improving the accuracy and performance of the face recognition model.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN XUMI YUNTU SPACE TECH CO LTD
- Filing Date
- 2022-10-20
- Publication Date
- 2026-07-17
AI Technical Summary
Existing face recognition models, when the training set is large, require the construction of a large number of sample pairs based on the sample pair learning method and rely on difficult examples. The prototype of the classification learning method cannot reflect the intra-class variance, resulting in poor face recognition performance and low accuracy.
By combining sample pair learning and prototype learning, the network weights of the face recognition model are updated by calculating the feature magnitude and quality score of the sample features and using the total quality score to weight the sample features and prototype.
It improves the accuracy and effectiveness of face recognition models, avoids the generation of large-scale sample pairs, and enhances the responsiveness of intra-class variance.
Smart Images
Figure CN115546867B_ABST