[2025 ICRA]RISED: Accurate and Efficient RGB-Colorized Mapping Using Image Selection and Point Cloud Densification

🛠 Getting Started

Prepare your data

Here we test this evaluation tool using dataset from Global-LVBA, you can download the data from link.

For the evaluation of geometric accuracy and surface coverage, preparing the global point cloud map and colorized point cloud map is sufficient. However, if you want to evaluate the projection accuracy, you're encouraged to prepare the initial datas of images and pcds bellow:

    RISED/
    └── data/
        └── sequence_name/
                ├── all_image/
                │   ├── 1661398632.022152.png     # image named by timestamp
                │   ├── 1661398632.121881.png
                │   ├── ...
                │   └── image_poses.txt           # camera poses (timestamp-aligned)
                ├── all_pcd_body/
                    ├── 1661398632.022152.pcd     # point cloud named by timestamp in body frame
                    ├── 1661398632.121881.pcd
                    ├── ...
                    └── lidar_poses.txt           # LiDAR poses (timestamp-aligned)

Intallation

Use conda to manage your Python environment:

conda create -n rised python=3.10 -y
conda activate rised
pip install -r requirements.txt

💡 Evaluation

Change the parameters in config/config.yaml.

To project and generate the global colorized map, you can run:

python3 utils/projection.py

For the three evaluation metrics and relevant visulization, you can run them seperately:

#  Projection accuracy
python3 example/evaluate_projection.py

# Geometric accuracy
python3 example/evaluate_geometry.py

# Surface coverage
python3 example/evaluate_projection.py

To evaluate all above three metrics in one manner and get the results precisely, run:

python3 main.py

📊 Visualization

Geometry Accuracy

Surface Coverage

Projection Accuracy

Photometric Error:

Valid Pixel Ratio:

🤗 Citation

If you find this repository useful, please use the following BibTeX entry for citation.

@inproceedings{rised,
  title={RISED: Accurate and Efficient RGB-Colorized Mapping Using Image Selection and Point Cloud Densification},
  author={Jiang, Changjian and Wang, Lijie and Wan, Zeyu and Gao, Ruilan and Wang, Yue and Xiong, Rong and Zhang, Yu},
  booktitle={2025 IEEE International Conference on Robotics and Automation (ICRA)},
  pages={3277--3283},
  year={2025},
  organization={IEEE}
}

👏 Acknowledgements

This repo benifits from Global-LVBA, LiDAR-VGGT.