Pacifiscan is an app created to help New Caledonians sort their waste. The project is structured as a French non-profit (Association Loi 1901, 501c equivalent). Currently, we have raised more than $5000 in grants.
How does it work ?
We wanted the app to be as intuitive as possible, so it is very simple. The user takes a picture of their waste, and the app uses Machine Learning to detect what type of waste it is and returns the nearest places to dispose of it.
The technical stack
Mobile app
Pacifiscan has two native mobile apps : iOS and Android. But because we didn’t had the ressources to work on two apps at the same time, I chose using React Native and Expo.
All the design and coding were done by me. You can download the app here:
http://go.pacifiscan.org/app
Image classification
To classify the image, we used CLIP from OpenAI. Having a zero-shot classification system left us a lot of free time for other tasks.
The API uses Flask and connects the CLIP model to an HTTPS endpoint.
Website
Pacifiscan is also available online here. It’s a NextJS app.
Design
All the design was done using Figma
Grant collection
We received two grants totaling 550 000 XPF ($5 025,11).
Pew Bertarelli Ocean Legacy
The first grant of 250 000 XPF ($2 284,14) was given by Pew Bertarelli Ocean Legacy (a joint project of The Pew Charitable trusts and the Bertarelli Foundation)
We defended our project to a jury and we won the grant.
Les nickels de l’initiative
The second grant of 300 000 XPF ($2 740,97) was given by Société Le Nickel, a subsidiary of Eramet.