---
tags:
- vision
language:
- en
pretty_name: Vis_diff
size_categories:
- 10K<n<100K
license: mit
task_categories:
- visual-question-answering
---
We process and collect entity and abstract concept. For emotion and visual style, we directly use [EMOSET](https://vcc.tech/EmoSet) and [Omni](https://huggingface.co/datasets/showlab/OmniConsistency).