TIDY offers offline semantic search for images on Android devices, enabling users to find photos using text descriptions or by providing a reference image. It targets Android users seeking private, efficient, and powerful image discovery capabilities without relying on cloud services.

How It Works

TIDY leverages a quantized version of the OpenAI CLIP model, pre-trained on the LAION-2B dataset. This approach allows for efficient on-device inference using ONNX Runtime. CLIP's ability to learn from text-image pairs enables semantic understanding for both text-to-image and image-to-image retrieval, providing accurate and generalized search results.

Quick Start & Requirements

• Install from Google Play Store.
• Requires an Android device.
• Initial indexing of the photo library is required on first launch.

Highlighted Details

• Offline operation ensures privacy and security.
• Supports both text-to-image and image-to-image search.
• Utilizes a quantized CLIP model for efficient on-device inference.
• Automatic indexing of new photos.

Maintenance & Community

• Project maintained by Viacheslav Barkov.
• No community links (Discord, Slack, etc.) are provided in the README.

Licensing & Compatibility

• The README does not explicitly state a license.
• Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

The initial indexing process may take time depending on the size of the photo library. The README does not specify the exact CLIP model variant or quantization details, which could impact performance and accuracy.