CVQA

CVQA

MH Specific

Culturally-diverse Multilingual Visual Question Answering benchmark with questions from 30 countries in 31 languages including Marathi. Images and questions are annotated by native speakers familiar with local culture

Build a culturally-aware Marathi visual question answering system that understands Indian visual contexts.
HomepageHuggingFace

Quick Start

from datasets import load_dataset
ds = load_dataset('cvqa/cvqa')
print(f'Samples: {len(ds["test"])}')
for ex in list(ds['test'])[:3]:
    print(f"Q: {ex['question'][:60]}")
    print(f"A: {ex['answer'][:60]}\n")
Modality
Image + Text
Size
~10K questions across 39 country-language pairs
License
Format
PNG/JPEG
Language
mr
Update Frequency
static
Organization
CVQA Benchmark Authors

Schema

FieldTypeDescription
imageimageImage for visual question answering
questionstringQuestion about the image in Marathi
answerstringAnswer to the question

Build With This

Create a multimodal Marathi educational tool that answers student questions about diagrams and charts
Develop a visual accessibility assistant that describes and answers questions about images in Marathi
Build a Marathi visual content moderation system that combines image understanding with text analysis

AI Use Cases

Visual question answeringcultural AImultimodal reasoning
Last verified: 2026-03-07