Sign in to view source links and access this dataset
Description
A building change detection dataset containing 15,000 pairs of satellite image patches. It was created by authors Zheng, Ermon, Kim, Zhang, and Zhong, with a related paper published in IEEE Transactions on Pattern Analysis and Machine Intelligence. The dataset page was last updated on October 10, 2025.
Use Cases
Training change detection models based on paired satellite imagery.
Benchmarking algorithms for identifying building construction or demolition.
Developing models to analyze urban development patterns from satellite time series.
Strengths
Contains 15,000 image pairs, providing a substantial volume for model training.
Focuses on two specific change types, offering clear classification targets.
Spatial resolution of 0.3-1 meter per pixel suggests high detail for building-level analysis.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is unknown, which may limit suitability assessment.
Data may reflect geographic or temporal bias inherent to its source collection.
Provenance
Source
EVER-Z on Hugging Face, associated with a GitHub repository and IEEE paper.
Collection Method
Likely derived from satellite imagery sources.
Freshness
Last updated 2025-10-10 18:54:22; freshness should be verified.
License is unknown; users must verify terms of use before application.