You are viewing the course site for a past offering of this course. The current offering may be found here.
Lecture 22: Image Processing (54)
knguyen0811

A description of the algorithm can be found here. The most interesting find on the page is that not only can the algorithm remove objects from an image but it can also reshuffle/move contents of images.

serser11

This reminds me of using deep learning for image completion. It's interesting in that it's interpreting images as samples from a probability distribution and using GAN in the process of completing the missing region. More can be found here: http://bamos.github.io/2016/08/09/deep-completion/#using-gz-to-produce-fake-images

kavimehta

Can GANs be used for more complex reconstructions? Say 3D textures or point clouds?

GKohavi

@serser11 To add on this, Nvidia has done some development on image inpainting using GANs that gets pretty decent results. More recently, they show that you can add semantic information to control what types of stuff the GAN fills the missing space in with. https://www.nvidia.com/research/inpainting/, https://nvlabs.github.io/SPADE/

You must be enrolled in the course to comment