Ed Chin gave a presentation on using neural networks for art restoration. He discussed using a generative adversarial network (GAN) to restore damaged images by training it on a dataset of artwork images. The GAN consists of a generator that produces restored images and a discriminator that evaluates them. It can restore large missing patches without needing predefined cost functions. Sample outputs showed the GAN realistically restoring damaged areas of images. Chin proposed a website where users could upload flawed images, mask damaged areas, and receive restored images from the trained GAN model. He concluded GANs have other applications like dimension reduction, super resolution, and video prediction.