A blogpost-sized rundown of stable diffusion's textual inversion, its disruptive impact in the art industry, and a test of a preliminary version of Mist, a cloaking tool that modifies art to ruin stable diffusion's models. My thoughts on it as someone who is both an aspiring data scientist and art hobbyist.
Click here to view the HTML webpage for this notebook
Note: In the notebook/guide I specify to clone this repository for the final version of the models. Due to storage limitations, I was only able to upload the vibes-concept-output model completely. vibes-misted-concept-output lacks its trained unet and vae models. You will have to download these models yourself.
Apart from that, the rest of the auxillary files for the notebook and the notebook itself are available.
This project was done in fulfillment of my individual requirements for my Machine Learning 3 class taught by Prof. Basti Ibanez.