AI image rating tool for Darktable – an open source post
This tool is designed to facilitate the analysis and evaluation of image collections through the use of local AI models.
What does the tool offer?
This project provides a comprehensive solution for everyone who deals with the management and analysis of large amounts of images:
- Automated image rating: The tool enables images to be evaluated through an artificial intelligence on a scale from 1 to 100. The AI provides an assessment based on configurable criteria.
- Detailed picture review: A constructive criticism is generated for each image, which illuminates various aspects such as exposure, focus or composition. This can serve as a starting point for your own image analysis.
- Automatic keyword generation: Relevant keywords or meta tags are created for each rated image. These can support the organization and later findability of the images.
- Wide format support: The tool handles a variety of common image formats, including RAW formats (e.g. .ARW, .CR2, .NEF) and .JPG, .JPEG, .TIFF and .TIF. An automatic conversion of RAW files is integrated.
- Flexible folder processing: A base folder can be selected and the processing can optionally be extended to subfolders, which simplifies batch processing of large image libraries.
- CSV export of results: All generated reviews, reviews and keywords are exported to structured CSV files. Both local CSV files are created per processed folder as well as a central overview.
- XMP sidecar file: XMP file to import into Darktable
- Customizable AI Configuration: The API parameters (e.g. API URL, System Prompt, Maximum Tokens, Temperature) and Image Settings (Width, Height, JPEG Quality) can be customized via a JSON configuration file to customize the tool to specific requirements.
- Web-based user interface: An intuitive web interface allows control of the evaluation process in the browser. Here, users can start and stop processing, track progress in real time and view detailed logs.
- Dark Fashion: A dark mode is integrated for eye-friendly use.
Operation
The tool is designed to work with local Large Language Models (LLMS) like LM Studio. It prepares images (including optional conversion and size adjustment) and sends them to the configured AI model. The AI analyzes the images and provides a structured answer with rating and keywords, which are then collected and stored in CSV files.
FAQ
Why the detour via an external program?
Using the Python application allows the use of the Darktable CLI. This is specially designed for batch processing and for this reason faster than the normal darktable version. The difference is clear, especially with large RAW files.