Mercyiris/remote-sensing-change-detection

Remote Sensing Change Detection Dataset

For Chinese documentation, please see README_zh.md

Dataset Description

A specialized dataset for remote sensing change detection research, containing complete image processing pipeline and annotation information. This dataset includes 24 groups of registered and aligned remote sensing image samples, with each group containing 5 different types of image files and corresponding annotation files.

Dataset Features

• Data Scale: 24 groups of image samples
• Image Types: Optical images, SAR images, binary change maps
• File Formats: TIF (original images), PNG (change binary maps), JSON (annotation files)
• Preprocessing Status: Registered, aligned, and cropped, not yet segmented by fixed resolution
• Annotation Completeness: Contains pixel-level change detection annotations

File Structure

The dataset contains the following 6 directories:

Image Files

• A/: Gaofen-2 pre-event optical images (.tif) - Reference baseline images for change detection
• B/: Gaofen-3 post-event SAR images (.tif) - Synthetic Aperture Radar images
• C/: Sentinel-2 unprocessed post-event optical images (.tif) - Raw optical images
• D/: Sentinel-2 relatively radiometrically corrected post-event optical images (.tif) - Preprocessed optical images
• E/: Binary change maps (.png) - Change detection results

Annotation Files

• json/: JSON annotation files corresponding to change maps, can be read and modified using LabelmeCD-AI

Dataset Applications

Primary Use Cases

1. Change Detection Algorithm Research - Develop and test new change detection methods
2. Multimodal Fusion - Research fusion techniques for optical and SAR images
3. Image Preprocessing Evaluation - Compare effects of different preprocessing methods
4. Deep Learning - Use as training and testing data

Research Directions

• Time-series remote sensing image analysis
• Multispectral image processing
• Urban building change monitoring

Technical Specifications

• Processing Status: Registered and aligned
• Channels: 3

Important Notes

1. File Integrity: Ensure consistent file counts across A, B, C, D, E directories
2. Preprocessing Requirements: Further resolution unification required based on specific application needs
3. Deduplication: Although each group of images is manually annotated separately, to avoid confusion between validation and training sets due to overlapping regions, deduplication can be performed based on coordinates

Citation

If you use this dataset in your research, please cite:

@dataset{remote_sensing_change_detection_2025,
  title={remote-sensing-change-detection},
  author={Tingxuan Yan},
  year={2025},
  publisher={Hugging Face},
  howpublished={\url{https://huggingface.co/datasets/Mercyiris/remote-sensing-change-detection}}
}

License

This dataset is released under the CC BY 4.0 license, allowing free use, modification, and distribution with proper attribution.

Contact

For any questions or suggestions, please contact:

• Hugging Face: @Mercyiris
• Email: [email protected]