Using All-in-One Perception Network (AOP-Net), a LIDAR based-multi-task framework that combines 3D object detection and panoptic segmentation Improving detection task performance with a 2D backbone designed to intertwine Multi-Layer Perceptron (MLP) and convolution layers How do we propose a model to guide the detection head by recovering useful features discarded during down-sampling operations in the 3D It is challenging to process LiDAR dataset due to its inherent geometry and massive volume. However, we will be exploring how we can utilise its great potentials of rich density and high spatial resolution, and use deep learning classes, such as Graph Neural Network based frameworks to learn identify the objects in the 3D Lidar point cloudsbackbone?