Browse outfit, comment and like
Change photo color with a big range, note face is also similar to yellow color but not changed
Change photo with a small range, note the bag color is not changed
Profile page, browse my outfits, likes and commets
Ideas (user suggested edits) to an outfit
Create an affiliation campign
Affiliation workers that pay affiliation as background job when user clicks
Upload photo with added affiliation url
Sharing life and decentralized affiliation system is lack on the current market. Traditional Affiliation program has centralized broker and control everything. But with Shinedme blockchain, everyone can be the provider themselves and make profit their own.
What it does
- Share fantastic outfit style with community
- Make different style by creating the same photo to exchange the aesthetics with each other.
- Boost your online store with decentralized affiliation.
How we built it
- Build a custom Kusama chain with customized pallets with substrate sdk and IPFS node
- Build auxiliary backend services to handle affiliation link redirection and serve as initial faucet
- Build a User-Friendly Web APP talk to blockchain and auxiliary backend services in React.js
- Implement a Open CV based image color changing editor with face recognition ML model.
Challenges we ran into
- Figure out a decentralized affiliation model that works with real world e-commerce websites and as part of pallet in chain
- Avoid changing color of the face in photos in Shinedme online image editor
- Integrate all services in a way that easy to setup and deploy
- Design a smooth on boarding workflow in frontend that is user friendly and not require much knowledge of blockchain
Accomplishments that we're proud of
- Great use experience since user can use the decentralized APP just like web2 APP and not make blockchain a difficult and complicated concept in the website. We have one click account creation process, easy to remember with 12-word mnenomics with local storage.
- Beautiful UI design to illustrate the theme of the website.
- Machine Learning model to help improve user experience on editing outfit photo styles by detect and cut face recolonization.
What we learned
- The culture and technology from Polkdadot blockchain and substrate sdk is great since it has highly composability, modularity and well documented.
- Decentralized system gives chance to people more control over their data and safe, low requirement to enter affiliation program and trust-less affiliation earning for social media users and easy secure setup for business user .
What's next for Shinedme
- Add more interesting ML modal for image editing features to make it even more creative and economically attractive to share an image, such as:
- clip an accessory or clothes from another image, and paste into the target photo to indicate "a better outfit style", and suggested user can earn tokens from purchase of suggested item replacement.
- change pattern of a cloth instead of color.
- Make it more secure and higher performance from current prototype, such as:
- replace usage of map of vectors to double maps.
- build indexer service that listening events from chain and organize data in a database to handle complex queries efficiently. Like the graph.com. Unfortunately there's no available library in Polkdadot Eco-system, if next time join hackathon we'd like to choose "tools" and build a the graph like indexer :) .
- allow user to backup generated key pair. Keys are not and will not stored on server to be decentralized, but it will be good to ask user to remember 12-word mnenomics and have recover account from 12-word feature.