Face Classification

From face image it is possible to determine a number of human attributes, such as gender, age, ethnicity, facial hair type and color, presence of spectacles, etc. It is also possible to determine face expressions, for example a smile.

The problem of visual appearance features estimation from face images is generally known as 'face classification'. Face classification is useful in a number of different application, e.g:

  • Demographic profiling of customers (e.g. for shop, cafe or retail bank office)
  • Measurement of efficiency and audience analysis of Digital Signage systems
  • Search of people in surveillance video archive
  • Intelligent human-computer interface

A pipeline of face classification algorithms is very similar to that of face recognition algorithms

  • Extraction of person's faces from an image (Face Detection)
  • Localization of facial keypoints (Face Features Detection). Often this stage is limited to specification of human's eyes only.
  • Face image normalization, including image resize to the predefined resolution, head inclination elimination, face color correction (Face Normalization)
  • Face descriptor computation. Contrary to the general knowledge, relative distances between facial keypoints are rarely used for face analysis and person identification. Usually a specific set of texture-based features are computed. (Feature extraction and descriptor computation)
  • Person attributes estimation from descriptor. Pattern recognition algorithms, based on machine learning, are usually used for it. (Classification)

The main performance characteristics of face classification algorithms are classification speed and precision.

Similarly to face recognition algorithms precision of face classification substantially depends on the input data quality: image resolution and sharpness, viewing angle, lighting conditions, person mimic. Depending on the image the precision of the same classification algorithm can decrease from 95% to 50-70%.

We have developed our own face classification algorithm. On our internal test dataset of 100K+ real-life photos gender classification accuracy is 90.1%, and mean age estimation error is 7.8 years.

To evaluate its performance on your application we offer to test our software on your image dataset. If you have a large dataset for your specific scenario, our face classification algorithm can be tuned (trained) on your data, which will guarantee its maximal precision on this scenario.

2010-2017 LLC "Video Analysis Technologies"