rabiulcste/vismin

[NeurIPS24] VisMin: Visual Minimal-Change Understanding

30
/ 100
Emerging

This project helps researchers and developers create specialized image and text datasets where an image and its description are subtly altered to test how well AI models understand nuanced changes. You input an original image and caption, along with instructions for a minimal change (like adding an object or altering a detail). The system then generates a new image and caption reflecting that small edit, which is useful for benchmarking visual AI models. This is ideal for those building or evaluating advanced AI for image understanding.

No commits in the last 6 months.

Use this if you need to generate high-quality datasets of 'minimal-change' image-text pairs to rigorously test and train AI models on their ability to detect subtle visual or textual differences.

Not ideal if you're looking for a general-purpose image editing tool or a solution for large-scale, complex image generation without a focus on minimal changes for AI evaluation.

AI-model-evaluation computer-vision-research dataset-generation multimodal-AI benchmark-creation
No License Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 6 / 25
Maturity 8 / 25
Community 16 / 25

How are scores calculated?

Stars

19

Forks

6

Language

Python

License

Last pushed

Mar 03, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/rabiulcste/vismin"

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