The framework works great with the pre-trained models the authors provide. A bit of Google Colab wizardry lets you set up dependencies, configure the model, and interact with raster color blobs that represent different parts of the room without looking at a single line of code. You can even manipulate blobs, see how they affect generated images, and export an
mp4 video of the input blogs and the resulting output image.
I look forward to learning more. A quick look at the paper and code repository felt overwhelming as it seems GANs are getting better and better as they increase in complexity and required knowledge to understand what they are really doing.
Have a great weekend!