Add dataset card for VLM-3R-DATA: paper links, abstract, tasks, usage, and architecture
#2
by nielsr HF Staff - opened
This PR initializes the dataset card for the Journey9ni/VLM-3R-DATA repository, providing crucial information for researchers and users.
It adds:
- A link to the associated paper: https://huggingface.co/papers/2505.20279
- Links to the project page: https://vlm-3r.github.io
- Links to the GitHub repository: https://github.com/VITA-Group/VLM-3R
- Appropriate
task_categories:video-text-to-text,question-answering - Relevant
languageandtagsfor better discoverability, including3d-reconstruction,monocular-video,spatial-reasoning,temporal-reasoning,embodied-ai,multimodal, andvision-language. - The full abstract of the paper.
- Key innovations of the VLM-3R framework.
- A detailed section on the VLM-3R architecture, including diagrams.
- A comprehensive "Datasets & Benchmarks" section, clarifying this repository's role as the main instruction tuning dataset.
- A "Sample Usage" section with code snippets for environment setup and model inference (video and image), extracted directly from the GitHub README.
- The BibTeX citation for the paper.