PointVisualizaiton

A point cloud visualization repo

• Multiple Point Cloud Renderer using Mitsuba 3
• Real Time Point Cloud View Tool

Dependencies

python >= 3.7

pip install -r requirements.txt

Quick Usage

# Render a single file to colorful image
python main.py --path example/chair.npy --render

# Render a single file to knn cluster colorful image
python main.py --path <file path> --render --knn

# Render a single file to white image
python main.py --path <file path> --render --white

# Render a single file to specified RGB
python main.py --path <file path> --render --RGB R G B

# Render a single file to split part using knn
python main.py --path <file path> --part

# Render a single file with rotation 90 degree in y axis
python main.py --path <file path> --render --rot 0 90 0

# Render a single file with voxelization style
python main.py --path <file path> --render --radius 0.03 --num 384 --type voxel

# view real time point cloud
python main.py --path <file path> --tool

# view real time colorful point cloud with bbox
python main.py --path example/furniture.npy --tool --bbox example/bbox.npy --BGR

Args Introduction

• Required Parameters

--path: Specify the path for the input file. Currently, supports .npy, .ply, .npz, .txt, and .pth formats for input, with a size of N × 3 (without color) or N × 6 (with color). If the size is B × N × 3, the first element in the batch will be selected.

--render: Using mitsuba to create beautiful image with shadow.

--tool: Using real time point cloud visualization tools, you can drag the point clouds. Typing "Q" to exit.

• Optional Parameters

--knn: Using KNN cluster to generate render color map. Note that KNN render will ignore the origin color infomation (if have).

--num: Specify the downsample point num, default is inf.

--center_num: The knn center num, default is 24.

--part: Perform KNN clustering on the objects and render each segment separately, with the 'center_num' parameter equally effective.

--white: Render white object. Note that white render will ignore the origin color infomation (if have).

--RGB: Render object with specific RGB value. Note that RGB render will ignore the origin color infomation (if have).

--mask: Render masked point cloud.

Source

Many thanks to following codes that help us a lot in building this codebase:

• PointFlowRenderer
• Mitsuba2PointCloudRenderer
• PointSetGeneration

Citation

If you find this project helpful for your research, please consider citing the following BibTeX entry.

@inproceedings{qi2023recon,
  title={Contrast with Reconstruct: Contrastive 3D Representation Learning Guided by Generative Pretraining},
  author={Qi, Zekun and Dong, Runpei and Fan, Guofan and Ge, Zheng and Zhang, Xiangyu and Ma, Kaisheng and Yi, Li},
  booktitle={International Conference on Machine Learning (ICML) },
  year={2023}
}
@inproceedings{dong2023act,
  title={Autoencoders as Cross-Modal Teachers: Can Pretrained 2D Image Transformers Help 3D Representation Learning?},
  author={Runpei Dong and Zekun Qi and Linfeng Zhang and Junbo Zhang and Jianjian Sun and Zheng Ge and Li Yi and Kaisheng Ma},
  booktitle={The Eleventh International Conference on Learning Representations (ICLR) },
  year={2023},
  url={https://openreview.net/forum?id=8Oun8ZUVe8N}
}