'Tax-free' 3DMM Conditional Face Generation
This addresses the trade-off between controllability and quality in face generation for computer vision applications, offering a solution to a known bottleneck.
The paper tackles the problem of lower sample quality in 3DMM conditioned face generation, showing that previous methods like DiscoFaceGAN and 3D-FM GAN have a significant FID gap compared to unconditional StyleGAN, and proposes a new model that effectively removes this quality tax.
3DMM conditioned face generation has gained traction due to its well-defined controllability; however, the trade-off is lower sample quality: Previous works such as DiscoFaceGAN and 3D-FM GAN show a significant FID gap compared to the unconditional StyleGAN, suggesting that there is a quality tax to pay for controllability. In this paper, we challenge the assumption that quality and controllability cannot coexist. To pinpoint the previous issues, we mathematically formalize the problem of 3DMM conditioned face generation. Then, we devise simple solutions to the problem under our proposed framework. This results in a new model that effectively removes the quality tax between 3DMM conditioned face GANs and the unconditional StyleGAN.