Skip to content

Dataset · JunDong-dev

Text-based object retrieval on satellite and SAR imagery

JunDong-dev/OptSAR-RSVG

A large-scale benchmark featuring optical and synthetic aperture radar (SAR) remote sensing images and text descriptions for visual grounding.

What's inside

The large-scale OptSAR-RSVG benchmark (from 10k to 100k examples) combines optical and synthetic aperture radar (SAR) earth observation images with English text descriptions and bounding box coordinates. The dataset is presented alongside the OptiSAR-Net++ framework, allowing a single model to retrieve objects by text query across two imaging domains.

What you can do with it

  • Training multimodal models for text-based object retrieval on satellite and SAR imagery
  • Benchmarking cross-domain detectors for remote sensing data
Tasks
Object detection
Languages
English
Size
10K–100K rows
Downloads this month
35.7K
Likes
1

How to load it

from datasets import load_dataset

ds = load_dataset("JunDong-dev/OptSAR-RSVG")