Skip to content

Dataset · Luckyjhg

Geo170K multimodal geometry dataset

Luckyjhg/Geo170K

Over 170k pairs of geometric images and texts for training multimodal models.

What's inside

The repository stores parquet data combining geometric figure images and text descriptions or QA pairs (metadata indicates 100k to 1M records, index card shows about 177k rows). Detailed assembly process and languages are missing from the dataset card itself.

What you can do with it

  • Fine-tuning vision-language models (VLM) for geometry problem solving
  • Training geometric drawing recognition and description
Size
100K–1M rows
Downloads this month
1.5K
Likes
46

How to load it

from datasets import load_dataset

ds = load_dataset("Luckyjhg/Geo170K")