matiasdelellis/facerecognition

What it solves

This project provides a facial recognition system for Nextcloud, allowing users to automatically organize and search for people in their photo collections without their images ever leaving their own server instance, ensuring privacy.

How it works

The application monitors for new image files and queues them for analysis. A background job (triggered by a schedule or an administrator) processes this queue using pretrained models and Dlib PHP bindings to detect faces and group them by similarity. Users can then name these identified groups (clusters) to make their photos searchable by person.

Who it’s for

Nextcloud administrators and users who want Google Photos-like facial recognition and person-based photo organization while maintaining full control over their data.

Highlights

• Privacy-focused: All image analysis happens locally on the Nextcloud instance.
• Automated Workflow: Automatically detects new images and queues them for analysis via background jobs.
• Searchable People: Integrates with Nextcloud's file and photo applications to allow searching for photos by person name.
• Parallel Processing: Supports multi-core machines via worker threads to speed up image analysis.
• Broad Format Support: Analyzes JPEG, PNG, and (depending on server configuration) HEIC/HEIF, TIFF, and AVIF.