It has been a bit since we have posted on this platform but we wanted to share some results and updates on our SEE Shell ML app that identifies tortoiseshell products from photos. The app takes a photo and compares it to our database of more than 5,000 images of tortoiseshell products and faux or similar looking products with an estimated accuracy of more than 90%.
The app has now been available for free on Apple & Android for about a year and a half (more info and links here).
We have had more than 1,000 people download the app and submit more than 3,500 images to date. We are partnering with organizations like WWF, Latin American Sea Turtles, and Fundacion Tortugas del Mar in Colombia, Indonesia, Costa Rica, Nicaragua, Panama, and elsewhere to test the app, train government and law enforcement officials, and educate the traveling public about this threat. We also had some great coverage of the app launch in National Geographic.
We've spent the last year analyzing the data, getting feedback from users, and upgrading different aspects of the app. Improvements include adding in filters to keep images that are not following instructions out (which improves accuray) and adding in a web component where users can access their submissions. Future improvements we hope to make are adding in modules for conservation practitioners to add in additional data and context for submissions, creating a group of data collectors to use the app in known hotspots on an ongoing basis, and updating our Global Tortoiseshell Report, which was last published in 2020.
We'd love to hear any feedback on the app or program from fellow conservationists and identify people that would be interested in helping to improve the app and collect data in the field.