Abstract:
Traffic density is turning into one of the major concerns issues around the world. Of
excessive traffic, people are facing health and mental problems. Around the world,
Bangladesh is 4th country that has the larger traffic. It is continuously increasingly timeconsuming,
causing fuel consumption, stress, delays, mental health, physical health, and
pollution. For improved signal management and efficient traffic control, it is also important
to determine real-time traffic density on the roads due to its ever-increasing nature. One
of the crucial components influencing traffic flow is the traffic controller. The result is that
it becomes necessary to optimize traffic management to meet this rising demand better.
Our suggested method intends to use real-time photos from the cameras at traffic
intersections for image processing and AI-based traffic density calculation. It also
emphasizes the algorithm for adjusting traffic signals depending on the number of
vehicles. We have collected around 1727 data locally, which we have trained. Our class
was Car, Bike, Bus, Truck, and CNG. We have used the YOLO3 model to solve this. And
it achieved the best accuracy of 86%. We also did a simulation to understand our model
better. Our top results are for all class precision of 86%, Recall of 0.741%, and mean
average of 0.805%.
