I used the KITTI dataset to train a 3D point cloud object detection model based on the Point Pillars architecture, focusing on urban environments. The model was trained to detect various objects such as cars, pedestrians, sidewalks, and other road elements. For 3D annotation, I utilized the Label Cloud tool and performed inference on publicly available datasets to validate the model’s performance. This project was a key component of my internship, and the resulting model was integrated into the autonomy software stack. It played a crucial role in enabling obstacle detection and the identification of environmental elements and objects within semi-urban environments.