Facial Recognition
One of the most interesting features of PhotoPrism is its AI-powered facial recognition. This lets you find pictures of your family and friends without any fuss or bother.
New faces are detected as you scan your library. They are then grouped by similarity.
To recognize faces, PhotoPrism uses a three stage process. First it extracts crops from your images using the Pigo face detection library (based on the Pixel Intensity Comparison-based Object detection paper). These are then fed into TensorFlow to compute 512-dimensional vectors for characterization. In the final step, the DBSCAN algorithm attempts to cluster these so-called face embeddings so that they can be assigned to people with a few clicks.
The face recognition accuracy is much higher than we expected.
Pages in this article:
Page 1 – Introduction / Installation
Page 2 – In Operation
Page 3 – Facial Recognition
Page 4 – Places
Page 5 – Summary
There’s a package for PhotoPrism in the AUR. Is that an easier install?
The AUR package does compile, but I haven’t tested it.
Steve, maybe worth adding PhotoPrism to the Google Photos alternatives?
Will do, but for now our Web Photo Gallery Solutions roundup has been updated. PhotoPrism is the winner!
Alternatives to Google Photos roundup now includes PhotoPrism