Skip to content

Dataset · initiacms

Visual Grounding Benchmark on Satellite Imagery

initiacms/XLRS-Bench_visual_grounding_en

English dataset with Earth imagery and object annotations for evaluating multimodal models in visual grounding.

What's inside

Part of the XLRS-Bench benchmark with a volume of 10,000 to 100,000 examples, combining images, English text, and time series. It is based on satellite and aerial imagery from DOTA (Google Earth and CycloMedia), ITCVD, and MiniFrance datasets with open annotations for non-commercial use.

What you can do with it

  • Evaluating multimodal LLM accuracy in object search and localization in remote sensing imagery
  • Testing computer vision models in satellite map analysis tasks
Languages
English
Size
10K–100K rows
Downloads this month
35.1K
Likes
6

How to load it

from datasets import load_dataset

ds = load_dataset("initiacms/XLRS-Bench_visual_grounding_en")