Meta-Album Airplanes: 800 Satellite Images of Aircraft Models

800 preprocessed satellite images of airplanes across 20 different aircraft types, sourced from Google Earth. The images were cropped and resized to 128x128 pixels by the Meta-Album team for few-shot learning benchmarks. The original dataset, MTARSI, was created by Zhize Wu and released under a Creative Commons Attribution license.

Use Cases

• Benchmarking few-shot image classification models based on the 20 aircraft classes.
• Training models for aircraft type recognition from satellite imagery.
• Developing data augmentation techniques for small, standardized image datasets.
• Researching domain adaptation for remote sensing imagery based on the preprocessed square format.

Strengths

• All 800 images are preprocessed to a consistent 128x128 pixel square format.
• Images were labeled by seven remote sensing specialists, suggesting high annotation quality.
• The dataset is explicitly designed for few-shot learning benchmarks within the Meta-Album collection.

Limitations

• The dataset is a downsampled micro version; the original contains over 9,000 images.
• Image resolutions were variable in the original source and have been altered by cropping and resizing.
• Column-level documentation and sample data are unavailable, requiring inspection after download.

Provenance

Source

Muti-type Aircraft of Remote Sensing Images (MTARSI) on Zenodo, sourced from Google Earth satellite imagery.

Collection Method

Images were cropped to squares and resized using an anti-aliasing filter by the Meta-Album team.

Geography

Images were acquired from around 36 airports worldwide.