#403 · Primary category: Computer Vision

tidy

android clip computer-vision cross-modal-retrieval deep-learning image-retrieval image-search image-text-matching image-text-retrieval kotlin nlp onnx quantization semantic-search

Offline semantic Text-to-Image and Image-to-Image search on Android powered by quantized state-of-the-art vision-language pretrained CLIP model and ONNX Runtime inference engine

Project last updated:03/28/24

GitHub Stars

597

Forks

45

Contributors

1

License

GPL-3.0

Why we included this project

TIDY gives your Android phone a real image search engine that works without a network. It runs a quantized CLIP model through ONNX Runtime entirely on-device, so you can type a description like 'sunset over the beach' or pick a reference photo and get visually similar matches from your own library. Because everything stays on the phone, your photos and the index built from them never leave the device, which is a genuine privacy win and also means the app works in areas with poor or no connectivity. The first launch scans your library and builds the index; after that, new photos are added automatically on subsequent opens, so the setup cost is one-time. That makes it a practical, self-contained way to do semantic image retrieval on Android, not just a model demo.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category