All's well that FID's well? Result quality and metric scores in GAN models for lip-sychronization tasks
arXiv:2212.13810v1h-index: 1
Originality Synthesis-oriented
AI Analysis
This work addresses lip-synchronization for video generation, but appears incremental as it adapts existing methods without clear novel breakthroughs.
The study tested GAN models for lip-synchronization by reimplementing LipGAN and comparing it to a variation called L1WGAN-GP, both trained on the GRID dataset, but no concrete results or numbers were provided in the abstract.
We test the performance of GAN models for lip-synchronization. For this, we reimplement LipGAN in Pytorch, train it on the dataset GRID and compare it to our own variation, L1WGAN-GP, adapted to the LipGAN architecture and also trained on GRID.