Neural network generated pixel monster NFTs
0xmons is an experimental NFT project featuring all unique pixel monsters created by a generative adversarial network (GAN). The animations are procedurally generated in Python with the GAN, and the names are outputted from a char-RNN. Human-curation is used to add the best ones to the blockchain. Users can either stake a token to earn them or purchase them. The motivation for this project was to create an NFT project with fully unique tokens, as well as exploring pixel art generation and animation. Currently, we've curated 90 of these unique NFTs. Later on, users will be able to merge two monsters to create a new monster and battle them, but it's a reach if we can push this out before the hackathon is finished.
How it's made
1) AngularJS + web3 for the front-end. Blocknative for simple wallet support. 2) Python, PyTorch, pandas, and Jupyter for the neural net + procedural animation stuff. The biggest breakthrough here was that we found that we could create interesting animations procedurally by adding noise to the input matrices before feeding them into the GAN. 3) Solidity + Truffle for the smart contracts. 4) Node.js + Heroku for the NFT metadata API.