Tencent Technology (Shenzhen) Co., Ltd. applied for a patent for a " training method, device, equipment, and storage medium for a facial forgery model ". According to the patent specification, with the development of artificial intelligence technology, its research and application in the field of the face (that is, face) counterfeiting are also increasing. At present, most face detection models based on deep learning rely on pseudo-face images with label information for supervised training. However, in practical application scenarios, there are limited pseudo-face images with label information, based on limited pseudo-face images. The identification accuracy of the face identification model obtained by image training is not high.
The present application discloses a training method, device, equipment, and storage medium for a face forgery model, and relates to the technical field of artificial intelligence. The method includes:
- Obtain a fake face image and a real face image, and the degree of difference between the facial pose of the fake face image and the face pose of the real face image is less than or equal to a threshold value;
- Based on the gradient information corresponding to the pseudo-face image, the pseudo-face image and the real face image are fused to obtain a fused face image;
- Based on the identification results corresponding to the fusion facial images, the face identification model is trained.
The patent abstract states that the embodiments of the present application can be applied to scenarios such as artificial intelligence, intelligent transportation, and assisted driving. Based on the fusion of facial images and label data, the self-supervised learning of the model is realized without being limited by the scarcity of samples, thereby improving the accuracy of the model's forgery detection.
0 Comments