Inspiration
Generative AI has had the ability to overtake the creative domain of artists, and NFTs minted on the blockchain created to help artists maintain control of their work have been violated by fraudsters. Additionally, the unauthorized minting of NFTs can lead to copyright issues with the artist's work. Therefore, we created VibeWire to be a secure platform where the use management and licensing of NFTs is handled through secure contracts and minting as well as our vector-based histogram model with high image comparison accuracy.
What it does
The front-end of our website enables you to enter your API key and begin minting. By using Pinecone vector-based database, we were able to build a model that can constructs a comparison histogram after comparing RGB pixels in images (to address unwanted replication, corruption, etc.). This model had an accuracy over 90% when we ran it. Our omnichain token minting saves the hash of the image so you and the artist/creator can ensure an image exists and is associated with an NFT.
How we built it
VerbWire: VerbWire was an offered track, and deployment of smart contracts was easy. Omnichain was a bit difficult, as some of our team members got processing errors. We also sometimes got processing errors when trying to deploy without adding an optional wallet ID. To avoid these issues, we created our own test wallets and deployed ERC-721 (standard NFT contract), as advised by VerbWire co-founder Justin when we met him during mentor office hours.
Three.js: Our teammate used Three.js in building the front-end for our website.
Pinecone, Numpy, and Pillow: Our teammate used Pinecone and these Python packages to help him build histograms that compared RGB values of images in our marketplace being traded, which helped in fractionalizing ownership of these image assets.
Challenges we ran into
One of the challenges we ran into was endless minting after an API key was entered on our front-end website portal. Another challenge was finding a functional GitHub repository we could add onto, as the localhost command was not working with some repos we came across despite our attempts to clear ports, remove firewalls, etc. (but we eventually found one!).
Accomplishments that we're proud of
We were able to develop a model that could do an image comparison histogram (the details of which are described above). This involved work with vector-based databases like Pinecone, and it had a high accuracy (90%+).
What we learned
We learned the power of VerbWire, deploying smart contracts, and NFTs. We learned how to work with image hashes as well as tying the front-end of our minting website to the back-end of our vector-based histogram model. Most of us entered as hackers with no prior use of VerbWire/the blockchain, and we all learned something new!
What's next for VibeWire
Let's continue to protect more assets of artists: images aren't the only form of art, and we want to help safeguard many more types of artists!
Log in or sign up for Devpost to join the conversation.