Nextcloud Recognize
AI media tagging app for your Nextcloud server. Groups photos by faces and labels objects, music genres and video actions. Open source under AGPL-3.0.
Nextcloud Recognize automatically categorizes media on your self-hosted Nextcloud server. It's for people who want to organize photo, video and music collections without sending sensitive data to a cloud recognition service. Image processing runs on your Nextcloud machine.
Face recognition groups photos by the people appearing in them, with a browsing interface in the Memories app. It also identifies objects such as animals, food, vehicles and buildings, plus landscapes, landmarks and monuments. For other media, it classifies music by genre and recognizes human actions in videos.
The app writes its results to Nextcloud's Collaborative Tags, so the categories become part of your existing media library. Memories displays tagged photos and videos, while the Audioplayer app lets you listen to tagged music. You can exclude folders from recognition without removing their files from Nextcloud.
Recognize uses pretrained models, including EfficientNet v2 for objects, Musicnn for music genres and MoViNet for video classification. TensorFlow.js handles image processing through a bundled Node.js runtime. The app needs roughly 4GB of free RAM. Native processing uses an x86-64 processor with AVX on Linux systems with glibc; a slower WebAssembly mode supports ARM processors and Alpine Linux, including Nextcloud AIO.
It's open source under AGPL-3.0. Maintenance focuses on compatibility and critical fixes, with community contributions accepted. Recognize can't process end-to-end encrypted files because the server cannot read their contents.
