Perceiving the environment, commonly known as Perception, is the core component of any Autonomous Vehicle (AV). It is achieved by connecting different sensors, such as Camera, LiDAR, RADAR, GPS, IMU etc. to the AV. Among all perception tasks, 3D Object Detection (3D-OD) is the most crucial one, as the AV needs to know the exact 3D geometry and position of different entities around its surroundings. It can be achieved with the use of the data obtained from a single sensor (Camera/LiDAR only approach) or fusion of data from multiple sensors (Sensor Fusion approach). In this work, we propose a LiDAR-only approach for 3D-OD using the point cloud data. This method eliminates the use of conventional expensive 3D convolutions to process the 3D point cloud data and instead uses only 2D convolutions for lighter processing while achieving comparable accuracy. The approach presents a pillar-based 3D-OD model with a cascade 2D backbone of ten(10) 2D convolution operations. This method can be deployed for real-time navigation tasks, which is a vital component in the context of an AV. The proposed model achieves decent performance on the KITTI benchmark test set for different standard accuracy measures, such as 3D detection, Bird’s Eye View (BEV) detection and Average Orientation Similarity (AOS). The results of our proposed model are listed on the leaderboard of the official KITTI website.