Meta-Album Boats: 1,040 Images of 26 Maritime Vessel Classes

A subset of the MARVEL dataset, this collection contains 1,040 square images of 26 different maritime vessel classes, each resized to 128x128 pixels. The dataset was preprocessed from the original MARVEL dataset by Dustin Carrion for the Meta-Album project in March 2022, with images adjusted to be square without cropping. It is intended for few-shot image classification tasks within the vehicles domain.

Use Cases

• Few-shot image classification based on 26 distinct maritime vessel classes.
• Benchmarking computer vision models on standardized, preprocessed 128x128 pixel images.
• Training object detection models for boats, as the description notes boats occupy most of the image frame.
• Studying model performance on vehicle imagery within a controlled, multi-class dataset.

Strengths

• Contains 1,040 images across 26 distinct classes, providing a multi-class classification benchmark.
• All images are preprocessed to a consistent 128x128 pixel square format, ensuring uniformity.
• Derived from the larger MARVEL dataset, which contains over 138,000 images, suggesting a substantive original source.

Limitations

• Description metadata is limited; actual data quality and class balance require manual inspection after download.
• Image resolution is low (128x128 pixels), which may limit detail for fine-grained classification tasks.
• The dataset is a mini subset; its scale of 1,040 images may be insufficient for training large models from scratch.

Provenance

Source

MARVEL: A Large-Scale Image Dataset for Maritime Vessels by Gundogdu, Solmaz, Yucesoy, and Koc.

Collection Method

Preprocessed (made square and resized) from the original MARVEL dataset for the Meta-Album benchmark collection.