Canopy is a privacy-preserving tool that helps conservationists sort camera trap data using machine learning without any data leaving the browser. It works best with datasets under 500 images. A test dataset can be found here.
Features:
- Powered by ONNX runtime
- Built-in caching. When using in the browser, downloaded models are stored in IndexedDB using localforage.
- No data leaves a users computer. Models can be deployed using only a few lines of code, meaning conservationists can finetune models locally and upload them to their browser easily.
- Canopy, uses Edge Functions to deploy models directly in web browser using open source packages like ONNX Runtime so no data leaves a users computer. Models can be deployed using only a few lines of code, meaning conservationists can finetune models locally and upload them to their browser easily.
Canopy has been built using Web AI, a TypeScript library that allows you to run modern deep learning models directly in the web browser. Special thanks to visheratin who built WebAI and has provided ad hoc assistance.
Mozilla Public License (MPL) version 2.0 (#MPL-2.0), which is both a free software license according to FSF and an open source license according to OSI.