Python | Virtual Environments

The False Promise of Virtual Environments

Just pip install -r requirements.txt, right? Right…?

Image by the author

I recently saw EditGAN, an artificial intelligence vision model that empowers the user to edit or manipulate the desired image with simple commands, such as drawing. The model works by first labelling parts of an input image. Then you can leverage the model to make edits to just that part of the image using text prompts. Some of the examples from the paper are the beak of a bird, the ears of a cat, or the wheels of a car. The wheels on your Toyota Prius too small? Select them and make them gigantic. Do you wish your cat had ears like a rabbit? Make it reality.

Papers With Code

Or maybe not. One of the coolest things about many of the research papers that are released by Nvidia, Google, and Meta is that they provide example code and pre-trained models. Or at least I thought it was pretty cool. This post was originally going to be a fun exploration of EditGAN, using AI to make my cat fat or a similar goofy task. This being one of my first experiences with a ML research project’s code, I went in with my hopes high, expecting to be playing with my cat pictures in a matter of minutes. After all, there was a requirements.txt file in the GitHub repository for…