cambridgeltl/visual-spatial-reasoning

[TACL'23] VSR: A probing benchmark for spatial undersranding of vision-language models.

38
/ 100
Emerging

This project offers a specialized dataset for evaluating how well AI models understand spatial relationships between objects in images. It takes image-caption pairs describing spatial arrangements (e.g., "The cat is behind the laptop") and checks if an AI can correctly determine if the description is true or false. Researchers and practitioners developing or benchmarking vision-language AI models would use this to precisely diagnose spatial reasoning capabilities.

140 stars. No commits in the last 6 months.

Use this if you need a focused way to test and compare how different vision-language AI models interpret and reason about spatial relationships like 'behind,' 'left of,' or 'at the edge of' within images.

Not ideal if you're looking for a broad benchmark that evaluates general visual reasoning, object recognition, or question answering, as this dataset specifically targets spatial understanding.

AI evaluation computer vision natural language understanding vision-language models spatial reasoning
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 12 / 25

How are scores calculated?

Stars

140

Forks

12

Language

Python

License

Apache-2.0

Last pushed

Mar 25, 2023

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/computer-vision/cambridgeltl/visual-spatial-reasoning"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.