In this study, we use CCTV and UAV video to detect and analyses the behaviors of small vehicles, such as motorbikes and autorickshaws, in under heterogeneous traffic, with a very high proportion of these vehicles, using the latest sensing technologies such as AI and deep learning. In addition to providing real-time information on traffic conditions through monitoring cameras, the system also helps in traffic management by reducing congestion and traffic accidents. Some of the researches are funded by JST and JICA as part of SATREPS project M2Smart "Development of Smart Cities in Emerging Countries through Multimodal Transport Systems Based on Sensing, Network and Big Data Analysis of Local Transport" (JPMJSA1606).