digiKam 8.6 update brings enhanced Face Recognition, Auto-Tagging, and more

digiKam 8.6 features a complete overhaul of the face recognition system, improving accuracy and speed by 25%-50% with full CPU utilization. The update includes a new face matching algorithm combining K Nearest Neighbor (KNN) and Support Vector Machine (SVM) classifiers, a revamped face detector that reduces false positives, and Face Image Quality Assessment (FIQA) to filter out low-quality images from training datasets. Image processing is now optimized by converting images to RGB, enhancing detection and recognition accuracy.

The user interface for face management has been improved with the use of YuNet and SFace for detection and feature extraction, now with added GPU processing support. The Auto-Tags Engine has been updated with new pipelines and improved classifiers, and an adjustable Auto Tagging confidence threshold has been introduced.

The "Image Quality Sorter" feature has been renamed to "Image Quality Scanner" for better usability and is now located in the Labels tab. Furthermore, red eye correction is now powered by a deep learning engine for enhanced precision. This release also includes numerous bug fixes and stability improvements that you can check here.