Detection of Lanes and Objects Using Deep Learning Techniques

Abstract

Over a period of time, the advancement in technology of computer vision has kept the users astonished. Computers replicate the processing of images same as that of humans and this branch of technology is called computer vision, which is part of artificial intelligence, and then provide the desired output. The goal of computer vision is to reconstruct the qualities of the world we observe. Recently, many experts and tech firms are in a race to prosper the development of self- driving cars. The ingenious innovation in the field of automation is self-driving cars. They provide a major solution to turn down the accidents and put a stop to damaging society. Self-driving cars have major issues in the detection of the road lanes and objects on the track. So, they need robust assistance to reach their destination safely. Consequently, sophisticated algorithms are used to detect lanes and objects appropriately. A deep learning approach is improved to obtain lane detection results. Our model implements a convolutional neural network (CNN) and OpenCV techniques to predict the lanes. Our model accepts the input video containing the road lanes and outputs the detected lanes stacked with green color. Subsequently, we extended the project to detect objects. To detect objects, we used the You Only Look Once (YOLO) algorithm that was trained using the COCO dataset. In the end, we combined the results of lane and object detection so that our final output will contain both detected lanes and objects.