Detection, Classification and Counting of Moving Vehicles from Videos

Detecting and Classifying vehicles into several categories demonstrates a considerable possibility in the field of intelligent transportation systems. In this paper, we aim to propose a Convolutional Neural network integrated with YOLO framework to detect, classify and count vehicles from videos considering different climatic conditions like rainy, foggy, night and sunny weathers. Experimental results show that on applying this proposed network on MIO vision Traffic Camera Dataset for training and testing the model with different traffic videos, the network has a good testing and classification ability taking into consideration of the complex features of the image dataset. Our algorithm achieves an average accuracy of 94.4% in the detection, classification and counting of vehicles depending on the visibility of the image in the frame.