DSTL

Inspiration

Tired of waiting for images to load on brokensea as you scroll through collections? I was, that's when I thought about how NFT collections have very constrained image characteristics that a neural network can easily learn and memorize.

What it does

DSTL trains a unique Deep Neural Network (autoencoder) model for your collection that allows you to hyper-compress your image payloads. Image payload sizes are around 10% of PNGs.

How we built it

• Train a deep autoencoder on the collection using pytorch.
• Mint an NFT on Polygon that points to our compressed image on IPFS
• Could potentially store images on-chain

Challenges we ran into

Training deep autoencoders

Accomplishments that we're proud of

Getting something working

What we learned

What's next for DSTL

Building infrastructure/API for training models and using the decoders Integrate the decoder framework into common libraries/languages

Built With

• material-ui
• python
• pytorch
• react
• solidity
• typescript