Implementation of Object Detection for Autonomous Vehicles by LiDAR and Camera Fusion

This paper focuses on the Object Detection technique that fuse 3D LiDAR points and 2D camera images. The fusion model uses ResNet 18 as a backbone for the Feature Pyramid Network (FPN) for object detection and distance estimation. The 3D LiDAR points are transformed on 2D Camera plane to predict the objects and the distance is evaluated from the LiDAR cloud points. Objects like cars, cyclists, and pedestrians are detected through boundary boxes by LiDAR and Camera Fusion. This network gives an accuracy of 98% for object detection and 97% for distance estimation for a LiDAR cloud point.