---
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.

## Acknowledgement
- [Japanese Visual Genome VQA dataset](https://github.com/yahoojapan/ja-vg-vqa)
- [DOCCI](https://huggingface.co/datasets/google/docci)

## Lisence
[Apach lisense 2.0](https://www.apache.org/licenses/LICENSE-2.0)