Datasets:

---

Multimedia-SMU

/

seeingculture-benchmark

like 3

Follow

Multimedia Lab @ SMU 7

This PR enhances the dataset card for the Seeing Culture Benchmark (SCB) by:

• Updating Metadata:
  • Changing task_categories from visual-question-answering and object-detection to image-text-to-text and image-segmentation to more accurately reflect the dataset's focus on VQA and cultural artifact segmentation.
  • Adding language: en and descriptive tags (vqa, cultural-understanding, cultural) for improved discoverability.
• Adding Top-level Links:
  • Inserting Markdown links to the paper (https://huggingface.co/papers/2509.16517), project page (https://seeingculture-benchmark.github.io/), and GitHub repository (https://github.com/buraksatar/SeeingCulture) at the top of the README for quick access.
  • Removing the outdated and commented-out HTML badge block.
• Adding Sample Usage:
  • Including a "Sample Usage" section with code snippets from the GitHub README to guide users on how to set up and use the MCQ Generator.
• Ensuring Link Consistency:
  • Correcting the GitHub link in the "Loading the dataset" section to https://github.com/buraksatar/SeeingCulture.
• Updating Citation Information:
  • Modifying the introductory text for the citation to reflect that the BibTeX information is already available below. The arXiv URL in the BibTeX itself remains unchanged as per guidelines.

These changes significantly improve the clarity, navigability, and completeness of the dataset card.