Deep Learning based Traffic Analysis of Motor Cycles in Urban City

This exploration expects to tackle the issue of traffic stream gauges utilizing information from a video surveillance camera. The target challenge is to count and categorize automobiles based on their traveling directions. This field is still in its infancy, and the focus of this research is merely one of the Trichy, Tamilnadu, India. This research work has utilized the cutting-edge YOLO v3 two-stage finder related to the SORT tracker to settle the expressed test. Vehicle movement direction was classified using a basic regions-based heuristic method. A few changes to the Faster R-standard CNN's presentation were made: center misfortune, versatile component pooling, additional veil branch, and anchors improvement. The proposed model plan and assess the marker and accumulated 982 video stream including over 60,000 things under organized situations. The preliminary findings reveal that during peak traffic periods, the suggested structure can count vehicles and detect their driving route with a mean overall rate inaccuracy of less than 10%. The dataset offered here is used by several scientists as a rigorous test or for truly preparing material.