metadata
license: apache-2.0
task_categories:
- image-to-text
language:
- ja
size_categories:
- 100K<n<1M
Dataset details
Dataset type: LLaVA JP Instruct 108K created by changing the data sets in Japanese Visual Genome VQA dataset and train data of docci_ja to the LLaVA-Instruct format.