VINBRAIN EXCELLENT RESEARCH ACCEPTED AT BRITISH MACHINE VISION CONFERENCE (BMVC) 2022!

VinBrain always strives to research and explore new technology methods to apply in life, promoting the country's digital transformation process. Therefore, VinBrain's staff has constantly built and developed researches in large, world-class conferences. VinBrain's excellent research was accepted at The British Machine Vision Conference (BMVC) in 2022 marks VinBrain's unremitting efforts in its journey to conquering technology.

The British Machine Vision Conference (BMVC) is an annual conference specializing in machine vision, image processing and pattern recognition. This is one of the major international conferences on computer vision and related fields, held in the UK. At the 2022 Conference, VinBrain's research was accepted and presented with a proposal to use deep learning networks to estimate the number of people present in the same place based on the Vision Transformer architecture.

The above research will make a great contribution to supporting people's lives and health, stemming from good control of crowd order and support for disease prevention. In 2021 and early 2022, realizing that monitoring and managing crowds through the security camera system is really necessary at a time when the COVID-19 epidemic has been complicated, it is necessary to take measures to disperse the crowds. Because of unusual crowds, limiting large gatherings, the team proposed using deep learning networks to estimate the number of people present in a location based on the Vision Transformer architecture.

In this solution, to estimate the number of people in the whole image, the team proposed a method to divide the image into many areas and estimate the number of people from each region, by extracting the feature of the resulting region match the correlation with spatial information in the vicinity and apply the Attention mechanism through the Vision Transformer model. Experiments show that the group's method is better than the previous methods when testing on standard data sets for the problem. 

The research team is very proud of this outcome and is expected to be contributed to the administration and maintain public order, especially at the time of the plague. In the meantime, VinBrain will continue to study and produce intelligent AI products, highly-funded to help improve the lives of the people!