Improve dataset card: Add description, links, tasks, and sample usage
#2
by
nielsr
HF Staff
- opened
This pull request significantly enhances the dataset card for Wakals/CoVT-Dataset by:
- Adding relevant
task_categories(image-segmentation,depth-estimation,image-to-image,image-feature-extraction,image-text-to-text),language: en, andtags(vlm,multimodal,visual-reasoning,chain-of-thought) to the metadata for better discoverability. - Providing a clear overview of the dataset's purpose, derived from the paper's abstract.
- Including direct links to the associated paper (Chain-of-Visual-Thought: Teaching VLMs to See and Think Better with Continuous Visual Tokens), the project page (https://wakalsprojectpage.github.io/comt-website), and the code repository (https://github.com/Wakals/CoMT).
- Adding a "Dataset Composition" section, including details and an image from the GitHub README.
- Incorporating a "Sample Usage" section with a basic Python snippet to demonstrate how to load and inspect the dataset using the Hugging Face
datasetslibrary, as hinted by the GitHub README's instruction to "Give it a shot!" on Hugging Face.
These additions provide comprehensive information for users interested in the CoVT-Dataset.