Face Detection and Recognition SDK
Software development kit for faces detection, facial recognition and analysis running on Windows and Linux
supports most common image formats as an input as well as live video streams or video files (SDK built in video capture or custom capture via grabframe callbacks)
multiple samples of usage, including complete projects with video support and persons database (C++, C#/WPF)
- multiple faces detection in images and video
- 22 basic facial points detection, face cropping
- 101 advanced facial points detection
- gender, age, smile, ethnicity detection
- glasses, skin color, facial hair, hair color type and hairstyle detection
- advanced faces recognition (search, identification, verification, faces similarity, 1:1, 1:N matching)
- RAM-cached very large faces search indexes support
- extended face geometry and colors measurements (face shape, facial features top-level description, relative sizes and locations), colors
- faces manipulation – morphing, transforming, mapping, and transition functions
- faces and facial features tracking in video
SDK overview and documentation
Recognition algorithm performance on LFW database
Mean classification accuracy: 0.9953
Standard error of the mean: 0.0009
Betaface API
Face detection and face recognition web service (HTTP)
You can upload images and retrieve image metadata, create face recognition databases and namespaces, search persons in your namespaces, verify identity and much more.
Free for non-commercial or low traffic commercial applications. Paid options include dedicated processing servers and licensing complete web service to run on your own hardware.
- multiple faces detection, cropped face images
- face landmarks locations (22 basic, 101 advanced)
- faces recognition and persons database
- gender, age, ethnicity, smile detection
- glasses, facial hair, hairstyle detection
- skin and hair color detection
- extended face geometrical measurements
- face features top level descriptors
- background color and clothes color detection
- 1:1, 1:N search, faces similarity