VQA v2 Dataset for Multimodal Question Answering

Over 100,000 entries combine images with question-answer pairs for visual question answering tasks. The dataset was created by lmms-lab and last updated in January 2024.

Use Cases

• Train a model to predict an answer from an image and a question text.
• Benchmark model performance on the visual question answering task using the provided image-question-answer triplets.
• Analyze the relationship between visual content and textual questions across over 100,000 examples.
• Develop models that jointly process image and text modalities for question answering.

Strengths

• Contains over 100,000 entries.
• Provides paired image and text data for multimodal tasks.

Limitations

• Specific row count, column names, and sample size details are unknown.
• Potential biases in image sources or question topics are not documented.

Provenance

Source

lmms-lab

Freshness

Last updated in January 2024.