Analyzing people’s social behavior with the use of images and videos is one of the most popular tasks for AI. Researchers have achieved a rather high quality in group-level emotion recognition, but until now it remained impossible to implement this development on a mass scale.
The problem was the requirement of most video systems for images containing face close-ups in good resolution. Ordinary cameras installed on the street or in a supermarket have resolutions too low and are mounted so high that the typical facial regions in the gathered videos are too small to work with.
However, this may no longer be the case.
Alexander Tarasov and Andrey Savchenko, researchers from HSE, have developed an algorithm that is comparable with the existing group-level emotion recognition techniques in terms of recognition accuracy (75.5%). At the same time, it requires only 5MB in the system memory, processes one image or video frame in just one hundredth of a second and can be used with low-quality video data.