Semantic Segmentation of LiDAR Point Cloud for Autonomous Vehicles

Course Project for CS541 - Deep Learning (Spring 2022)

Master of Science in Robotics Engineering at Worcester Polytechnic Institute

Team Members

• Chinmay Madhukar Todankar
• Bhushan Ashok Rane
• Aniket Manish Patil

Project Description

Dependencies

• NVIDIA CUDA Toolkit
• NVIDIA CuDNN
• Tensorflow
• NumPy
• Matplotlib
• Pillow

Dataset - Semantic KITTI

Download the Velodyne Sensor data and Label data, and place in the dataset folder in the form as mentioned on the Semantic KITTI website.

• Download Point Cloud Data
• Download Labels Data -> Save this link -> http://www.semantic-kitti.org/assets/data_odometry_labels.zip

The path of this dataset folder will be required as a argument to the run the entire pipeline.

Usage Guidelines:

Go to the parent folder of this repo, that is, semantic_segmentation and enter the command:

python3 scripts/main.py -d **path_to_dataset_folder**

References

1. A. Milioto and I. Vizzo and J. Behley and C. Stachniss, RangeNet++: Fast and Accurate LiDAR Semantic Segmentation