Autonomous Vehicle Image Classification using Deep Learning
P. Ramakrishnan, K Dhanavel, K Deepak, R Dhinakaran · 2023
High-performance computer systems, Deep learning algorithms, and sensors work together to enable autonomous driving. First, information about the surroundings around the vehicle is gathered via sensors like lidar, radar, and cameras. Then, deep learning algorithms evaluate this data to determine how the car should react, such as with steering or braking. These computations are done in real-time using high-performance computer devices There is ongoing research towards the creation of autonomous vehicle image categorization systems, however the systems confront several difficulties. In recent years, several approaches—including the application of deep learning algorithms and the incorporation of sensor fusion technology have been suggested to solve these problems. The precise and effective categorization of photographs taken in actual surroundings continues to present difficulties. The requirement to analyse massive volumes of data in real-time, which necessitates powerful computation and effective algorithms, is one of the key obstacles. In this paper we include the idea of detecting vehicles and things on the road, such as vehicles and lane detection, which can be used for autonomous vehicles. The main objective to achieve this our proposed system includes a popular algorithm YOLO (You Only Look Once) YOLOV3. By including the concept of detecting objects using YOLOV3, better results can be obtained by increasing the grid size to $26*26$. This extended grid size helps to detect the smaller objects.