In our research, we further develop our end to end system that processes video streams in real-time using deep learning, pattern recognition and many other methodologies to monitor and count pedestrians in road traffic environments. The developed AI system uses regular traffic cameras coming from the city of Los Angeles to control and analyze real-time traffic videos with greater focus on pedestrian traffic.There are a number of challenges presented when using these traffic cameras such as trajectory misgeneration and pedestrian misclassification. The objective is to fine-tune our pedestrian tracking and trajectory prediction to monitor pedestrian traffic and observe their travelling patterns. We also propose a transfer learning-based model to improve pedestrian recognition and expand its capability to identify more traffic-related objects. The system introduces a new lane detection feature to better manage traffic flow and gain more information of a vehicle's position and direction of movement.California State University, Los Angeles has partnered with leading industry clients the City of Los Angeles, the Los Angeles Department of Transportation (LADOT), California Department of Transportation (CalTrans), and Toyota Mobility Foundation to develop this highly advanced, modern system for real-time monitoring, tracking and counting of pedestrians.