CLIP-Finder v1.3

CLIP-Finder enables semantic offline searches of images from gallery photos using natural language descriptions or the camera, with predictions around 1ms using the Neural Engine.

Changes

• Added the CLIP-Finder submodule in the CLIP-Finder directory.
• Added CLIP-Finder description to the README.md.

Motivation and Context

Inspired by the examples in coreml-examples, I would like to include this project. Adding this submodule will enhance the repository with advanced image and text similarity search capabilities based on MobileCLIP. (According to the authors, CLIP Image is based on FastViT), leveraging CoreML and on-device processing for efficient and private operations.

🚀 Performance Enhancements

• 🏎️ Batch Processing: Implementation of batch processing to optimize the preprocessing stage during the first launch of the app. This significantly reduces the time required to index the photo library.
• 🔄 Sync Button was added, enabling synchronization with the Photo Gallery without restarting the app. It updates with new photos and removes deleted ones from the records.

Checklist

• I have tested these changes locally.
• I have updated the documentation as needed.
• Release available on TestFlight.