Exploring the utilization of rich 3D point clouds and leveraging the inherent spatial data within these clouds to enhance object detection and classification systems, unlike traditional image-based methods, which often struggle in complex environments. The use of LIDAR-derived 3D point clouds offers a robust foundation for creating accurate and reliable classification systems. The approach focuses on efficient feature extraction and modeling from point cloud information, for many applications like autonomous vehicles and robotics. The use of 3D point clouds enables comprehensive spatial analysis, allowing the system to perceive objects from multiple viewpoints. This helps in classification of the objects in the real world of streaming data (in each frame of video) in BEV (Birds Eye View) image, which in turn reflects in RGB image. This detection system uses different color bounding boxes for classification. For this it uses one of the most advanced model architecture ResNet18 with FPN. This advancement promises to revolutionize the capabilities of autonomous systems, contributing to safer and more efficient operations in complex environments.