GFP-GAN, or Generative Facial Prior, is an artificial intelligence-powered tool that can restore and enhance old images. The source image can be low quality and still generate impressive results. How so? It doesn’t simply use a pre-trained StyleGAN and then re-train it to orient the encoded information for their task as PULSE does.
That’s right, it utilizes a pre-trained StyleGAN-2 model to orient their own generative model at multiple scales during the encoding of the image, right down to the latent code and up to reconstruction. Simply put as possible, instead of orientating the training on just the generated (fake) image and the expected (real) image using a discriminator model from the GAN network, GFP-GAN also has two metrics for preserving identity as well as facial components. Speaking of identity, AI can also show us what cartoon characters look like as real-life humans.
- [Small but Powerful]-Mini PC about 5.04 x 5.04 x 2.05-Inch. Powered by Intel Celeron Jasper Lake N5105(4MB cache, up to 2.9Ghz), delivers incredible...
- [Intelligent Design]-What makes the small computer stand out from other mini PCs is that it has an empty expansion bay to allow adding a 2.5-inch SSD...
- [Ready When You Are]-Mini computer was pre-installed Windows 11 Pro. With the Modern light feature, blue light while running, red light while standby....
The facial component loss is basically the same thing as the discriminator adversarial loss we find in classic GANs but focuses on important local features of the resulting image like eyes and mouth. The identity preserving loss uses a pre-trained face recognition model to capture the most important facial features and compare them to the real image to see if we still have the same person in the generated image,” said Louis Bouchard.
