reonokiy/vsp-encyclopedic-vqa

---
license: apache-2.0
task_categories:
- visual-question-answering
- information-retrieval
tags:
- vision-self-play
- multimodal
- knowledge-intensive
---

# Encyclopedic-VQA (Vision Self-Play format)

Encyclopedic visual QA (Google, ICCV 2023), converted to unified Parquet schema.

> Mensink et al., Encyclopedic VQA: Visual questions about detailed properties of fine-grained categories. ICCV 2023.

## Schema

| Field | Type |
|-------|------|
| wikipedia_title | large_string |
| wikipedia_url | large_string |
| question_original | large_string |
| question | large_string |
| question_type | large_string |
| answer | large_string |
| evidence | large_string |
| evidence_section_id | large_string |
| evidence_section_title | large_string |
| dataset_name | large_string |
| dataset_category_id | large_string |
| wikipedia_url_used_in_train | large_string |
| encyclopedic_vqa_split | large_string |
| dataset_image_ids | large_string |

## Splits

| Split | Rows | Size |
|-------|------|------|
| test | 5,750 | 0.5 MB |
| train | 212,338 | 14.6 MB |
| val | 2,950 | 0.3 MB |
| **Total** | **221,038** | |

## Usage

```python
from datasets import load_dataset

ds = load_dataset("reonokiy/vsp-encyclopedic-vqa")

# 或加载特定 split
train = load_dataset("reonokiy/vsp-encyclopedic-vqa", split="train")
```