Automatic activity recognition

Automatic recognition of human activities from videos is already a reality. Using neural networks and computer vision capabilities, researcher Carlos Antônio Caetano Júnior has developed a new approach to optimize this process in his latest article.

The differentiator of the researcher’s work is the use of the Magnitude-Orientation Stream (MOS) technique to learn the movement in a more detailed way and thus predict the activity performed on the video.

The example above shows examples of activities involving interactions between two people: they are at different depths during such activities, and yet the algorithm recognizes accordingly.

Thus, this model has been shown to be more efficient than classical literature approaches that make use of local characteristics and other techniques of recent neural networks, thus demonstrating great potential for application.

Thus, the MOS approach has considerable benefits for developing intelligent systems that require a high level of security, such as surveillance systems that can use this technology to prevent abnormal activity, such as suspected theft and hijacking.

 

Journal of Visual Communication and Image Representation, Volume 63

https://www.sciencedirect.com/science/article/pii/S1047320319302172?dgcid=author

 

Carlos Antonio Caetano Júnior
PhD in Computer Science at Federal University of Minas Gerais (UFMG) and researcher at Smart Surveillance Interest Group – SSIG / DCC / ICEx / UFMG. Developed part of the doctoral studies at the INRIA Sophia Antipolis Recherche Center, France (CNPq scholarship), as a researcher on the STARS team (under the guidance of Dr. François Brémond).