Skip to content

Dataset · Voxel51

GQA sample of 35k images with scene graphs

Voxel51/GQA-Scene-Graph

35,000 images from the GQA dataset in FiftyOne format with object and attribute scene graph annotations, but without questions.

What's inside

This is a subset of 35,000 samples based on the well-known GQA (Visual Reasoning in the Real World) dataset, prepared for the FiftyOne framework. It contains images and scene graph annotations in English, structured as object-level detection attributes. Unlike the original GQA, the question-answer pairs themselves are not included in this version.

What you can do with it

  • Training and testing object detection models with attribute prediction
  • Scene graph generation from real photographs
Tasks
Object detection
Languages
English
Size
10K–100K rows
Downloads this month
49.5K
Likes
4

How to load it

from datasets import load_dataset

ds = load_dataset("Voxel51/GQA-Scene-Graph")