Fast Face Image Synthesis with Minimal Training
This addresses the need for privacy-safe face image datasets for training CNNs and biometric verification, though it is incremental as it builds on existing component-based and 3D rendering techniques.
The paper tackles the problem of generating realistic face images for data augmentation and biometric testing by sampling face components from real images and rendering them with a 3D head model. The result is a method that produces synthetic faces with different attributes and is compared to GANs in terms of quality and speed.
We propose an algorithm to generate realistic face images of both real and synthetic identities (people who do not exist) with different facial yaw, shape and resolution.The synthesized images can be used to augment datasets to train CNNs or as massive distractor sets for biometric verification experiments without any privacy concerns. Additionally, law enforcement can make use of this technique to train forensic experts to recognize faces. Our method samples face components from a pool of multiple face images of real identities to generate the synthetic texture. Then, a real 3D head model compatible to the generated texture is used to render it under different facial yaw transformations. We perform multiple quantitative experiments to assess the effectiveness of our synthesis procedure in CNN training and its potential use to generate distractor face images. Additionally, we compare our method with popular GAN models in terms of visual quality and execution time.