As per usual, read through the papers and make sure you understand them
Download and train a model using the full imagenet 1000k dataset
Try improving one of your extra generative models by adding WGAN on top of it
Use WGAN to create a generative model from any earlier paper, such as colorization, 3d view changing, noise removal, depth prediction, etc
In addition, next week we’ll be taking a look at mean shift clustering. So if you’re interested, see if you can implement it yourself in tensorflow or pytorch.
Post below with any questions, comments, or ideas.
Can’t get wgan-pytorch notebook to run on my box (used to run on AWS P2 instance just fine). @jeremy helped me debug the DCGAN notebook which does work properly now - suggesting the issue might be with pytorch for the wgan. Installed using the packages using instructions from pytorch.org, so not sure what issue might be. Googled for answer but did not find a solution specific to my setup
Error copied below. Appreciate any suggestions. cc: @rachel
