The current situation of traffic in Vietnam has many outstanding problems, especially traffic congestion, since the supply of infrastructure has often not been able to keep up with the growth in mobility. Thus, proposing monitoring plans to support authorities to make suitable and prompt decisions has always received large attention from the community. Meanwhile, applying information technology, especially advanced models which could process or analyze traffic data in real time is recently considered to be a priority solution due to the time, accuracy, and cost saving that it can potentially achieve. Therefore, this paper outlines research on three advanced real-time object detection methods: YOLOX, YOLOF, and YOLACT and the development of the newest Vietnamese traffic dataset named UIT-VinaDeveS22. The work contains both theoretical and empirical analysis,  which are expected to create premises for further studies into addressing problems such as traffic density management, traffic separation, and traffic congestion.