harrylal / simulation-of-birds-eye-view-map-generation-from-rgbd-data

Project: Generating overhead birds-eye-view occupancy grid map with semantic information from lidar and camera data.

Simulation of Birds Eye View Map Generation from RGB-D Data

This project aims to employ sensor fusion techniques utilizing data from lidar and camera sensors within the KITTI dataset to generate an informative birds-eye-view occupancy grid map enriched with semantic details

Key Features • Download • How To Use • Credits • License

Key Features

1. Bird's Eye View Occupancy Grid Map Generation: Creation of a comprehensive occupancy grid map from LiDAR data providing a top-down view.

2. Semantic Segmentation Mask Prediction using BiSeNetv2: Employing BiSeNetv2 for predicting semantic segmentation masks of the identified objects.

3. Point Cloud Projection onto Image for Semantic Information: Projection of the point cloud overlapping the camera's field of view onto images to extract semantic information about objects.

4. DBSCAN-Based Object Clustering in Point Cloud: Implementation of DBSCAN for clustering objects within the point cloud data.

5. Temporal Tracking of Objects for Semantic Identity Retention: Tracking the identified objects in a temporal domain, ensuring semantic identity retention even if objects move out of the camera's field of view.

Download

1. Download KITTI data files and caliberation files KITTI
2. Download BiSeNet Weights Drive

How To Use

1. Set the path to the downloaded KITTI dataset files in the config/settings.yaml file.
2. Set Bisenet Weights Path in config/settings.yaml file.
3. Execute the main.py file to start the project.

   python3 main.py

Credits

• BiSeNet
• KITTI
• PointPainting

License

This project is licensed under the MIT License - see the LICENSE file for details.