Meta-Album APL Mini: 840 Satellite Images of 21 Aircraft Types

840 remote sensing images of airplanes, resized to 128x128 pixels, form this dataset. It contains 21 different aircraft types, originally sourced from Google Earth satellite imagery and labeled by remote sensing specialists. The dataset was created by Philip Boser for the Meta-Album project in March 2022.

Use Cases

• Benchmarking few-shot learning algorithms based on the 21 distinct aircraft classes.
• Training image classifiers for aircraft type recognition from satellite imagery.
• Developing models for remote sensing object detection using pre-processed, square-cropped images.
• Studying domain adaptation for vehicle recognition across different visual domains.

Strengths

• Contains 840 images across 21 distinct aircraft classes, providing a multi-class benchmark.
• Images were carefully labeled by seven remote sensing specialists, suggesting high annotation quality.
• All images are pre-processed to a uniform 128x128 pixel size, ensuring consistency for model input.

Limitations

• The dataset is a mini version; the original contains over 9,000 images, so this is a subset.
• Image resolution is variable in the original source and has been downscaled, which may limit detail for fine-grained classification.
• Column-level documentation and sample data are unavailable, requiring inspection after download to understand data structure.

Provenance

Source

Muti-type Aircraft of Remote Sensing Images (MTARSI) dataset on Zenodo, created by Zhize Wu.

Collection Method

Images were acquired from Google Earth satellite imagery, cropped to be square, and resized.

Geography

Images were acquired from around 36 airports worldwide.