Improve dataset card: Add paper, project page, code, task categories, tags, and sample usage
#4
by nielsr HF Staff - opened
This PR significantly enhances the dataset card for the opendwm-data repository.
Key changes include:
- Linking the dataset to its associated paper: CVD-STORM: Cross-View Video Diffusion with Spatial-Temporal Reconstruction Model for Autonomous Driving.
- Adding links to the project page (https://sensetime-fvg.github.io/CVD-STORM/) and the GitHub repository (https://github.com/SenseTime-FVG/OpenDWM) for easy access to more information and code.
- Including relevant
task_categories(text-to-video,text-to-3d) andtags(autonomous-driving,video-generation,3d-reconstruction,diffusion-models,lidar) to improve discoverability on the Hugging Face Hub. - Expanding the dataset description to clarify its purpose within the OpenDWM project and its relation to 4D scene reconstruction for autonomous driving.
- Listing examples of specific data packages (like
nuscenes_scene-0627_package.zipandnuscenes_scene-0627_lidar_package.zip) that are referenced in the OpenDWM GitHub repository's usage examples, helping users understand the content. - Providing a comprehensive "Sample Usage" section, including environment setup and code snippets directly from the GitHub README, demonstrating how to use these data packages for layout-conditioned video and LiDAR generation.
- Adding a "Citation" section for the OpenDWM project.
These updates provide a much more comprehensive, informative, and discoverable resource for researchers and practitioners.