Good AI needs good data. Good data is expensive. By tying the collection of data to the actual service that is produced by the AI, it's possible to create very good AIs for very cheap.

fashion gallery is a proof of concept of this idea: a search engine for fashion e-commerces that uses AI to make rough predictions of what a product is about, then uses user data to improve the predictions. The search engine is continuously improving itself thanks to the AI backing it up, increasing the value of the search engine for the merchants.

How it works

In this proof of concept, the merchant is expected to add a bunch of images into the image/ folder of the project and then run the commands:

flask init-db
flask classify-images

These commands will send the images to Samasource's image fragmentation API and the save the results in a SQLite database.

That data is then used by a Python server to create the search engine.

How I built it

Python for the backend server. React for the frontend.

What's next for fashion gallery

Turning this proof of concept into a Shopify App would make a lot of sense.

Share this project: