EuroSAT Dataset
This dataset contains Sentinel-2 satellite images, in RGB and multispectral, classified according to different types of land use (forests, crops, urban areas, water, etc.). Each image is 64x64 pixels with a spatial resolution of 10 meters.
Approximately 54,600 64x64 pixel images, JPG (RGB) and multispectral TIFF formats
CC0: Public Domain
Description
The EuroSAT Dataset includes more than 54,000 Sentinel-2 satellite images, divided into land use classes such as annual crops, forests, industrial areas, roads, and bodies of water. The dataset includes RGB images as well as multispectral images in TIFF format, allowing detailed remote sensing analyses.
What is this dataset for?
- Forming models for the classification of land uses in agriculture and the environment
- Develop tools for geospatial analysis and monitoring of territorial changes
- Testing computer vision algorithms on multispectral images
Can it be enriched or improved?
This dataset can be enriched by adding new recent images or by additional annotations (e.g. segmentation, object detection). Methods for increasing multispectral images are also possible to improve the diversity of data.
🔎 In summary
🧠 Recommended for
- Remote Sensing Researchers
- Geoinformatics students
- Environmental AI developers
🔧 Compatible tools
- QGIS
- TensorFlow
- PyTorch
- Rasterio
- OpenCV
💡 Tip
Combine multispectral bands to extract plant indices (NDVI) to improve classification.
Frequently Asked Questions
What is the spatial resolution of the images in this dataset?
The images have a spatial resolution of 10 meters per pixel, with a size of 64x64 pixels.
Does this dataset contain multispectral images?
Yes, it contains RGB images as well as multispectral TIFF files with multiple spectrum bands.
Can this dataset be used for the segmentation of satellite images?
Yes, although this dataset is primarily annotated for classification, it can be adapted for segmentation tasks with additional annotations.
