Omni-CAD Dataset

The Omni-CAD dataset was introduced in the paper CAD-MLLM: Unifying Multimodality-Conditioned CAD Generation With MLLM.

Omni-CAD is presented as the first multimodal CAD dataset, designed to facilitate the training of unified Computer-Aided Design (CAD) generation systems. It enables the creation of parametric CAD models conditioned on diverse multimodal inputs, including textual descriptions, images, and point clouds. The dataset contains approximately 450K instances, each equipped with corresponding multimodal data, such as textual descriptions, multi-view images, point clouds, and CAD command sequences.

• Project Page: https://cad-mllm.github.io/
• GitHub Repository: https://github.com/CAD-MLLM/CAD-MLLM

Dataset Download and Usage

The Omni-CAD dataset can be downloaded using wget as shown below. For detailed preprocessing guidance, environment setup, and further usage (e.g., mesh export, point cloud sampling, image rendering), please refer to the official GitHub repository.

mkdir data
cd ./data
wget https://huggingface.co/datasets/jingwei-xu-00/Omni-CAD/resolve/main/Omni-CAD.zip
unzip Omni-CAD.zip
cd ..

The expected directory structure after downloading and unzipping is:

└── data
    └── Omni-CAD
        └── json            # CAD model json descriptions
            ├── 0000
            ├── 0001
            └── ...
        └── txt             # Text captions
            ├── 0000.json
            ├── 0001.json
            └── ...

Further preprocessing steps described in the GitHub repository will add step (or ply), pcd, and img directories.

Citation

If you use the Omni-CAD dataset in your research, please cite the following paper:

@misc{xu2024CADMLLM,
      title={CAD-MLLM: Unifying Multimodality-Conditioned CAD Generation With MLLM}, 
      author={Jingwei Xu and Chenyu Wang and Zibo Zhao and Wen Liu and Yi Ma and Shenghua Gao},
      year={2024},
      eprint={2411.04954},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}