Dataset · google
Google Remote Sensing Change Captioning Benchmark
google/RSRCC
A Google Research dataset with multi-temporal aerospace image pairs and questions about terrain changes.
What's inside
The RSRCC benchmark from Google Research's Remote Sensing Foundation Models team contains between 100,000 and 1 million English examples. It features multi-temporal satellite and aerial image pairs paired with natural language questions and answers describing semantic terrain changes. Suitable for VQA and multiple-choice tasks.
What you can do with it
- Evaluating and fine-tuning Vision-Language Models (VLMs) for satellite image analysis
- Training automated monitoring systems for urban development, landscape, and terrain changes from photos
- Benchmarking geospatial foundation models in visual-question-answering tasks
- Tasks
- Visual question answering, Image text → text, Multiple choice
- Languages
- English
- Size
- 100K–1M rows
- Downloads this month
- 16.2K
- Likes
- 41
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("google/RSRCC")