BitRefine: How many faces can you simultaneously capture from single Full-HD frame?

Source: Deep Learning on Medium

BitRefine group is constantly working on improving recognition capabilities of its multi-purpose deep visual analysis platform “BitRefine Heads”. One of the modules — is a face detection and recognition module. Now it is capable of reliably capturing more than 40 faces from single 2Mpix frame of real-world footage. Here is example:

This short GIF gives in idea of scene’s dynamics and if you’d like explore details, here’re full resolution screenshots:

The code in each frame is a unique tracking label. All recognized faces get pulled out of raw footage, recognized, counted and presented in form of statistical charts and tables. Surely if you need to find someone it supports searching.

BitRefine Heads is a new generation of computer vision software that combine traditional recognition algorithms with almost infinite capabilities of artificial “brain” based on deep neural networks. Recognition algorithms are based on artificial neural network structures that allow recognizing not just faces, but almost every object that human can recognize. Please see this article.

At the moment is actively developing international markets and looking for new partners. If you’re interested in advanced computer vision solutions, please contact us via official website or just send “Hello” to


BitRefine group is a data science and computer vision. We specialize in deep learning video detection and help companies solve the cases where standard detectors and approaches don’t work.