Dataset Overview

Dataset Description

Fine‑Grained Visual Classification of Aircraft (FGVC‑Aircraft) is a benchmark dataset for fine‑grained visual classification of aircraft.

• Language (NLP): English
• License: Other

Dataset Source

• Homepage: https://www.robots.ox.ac.uk/~vgg/data/fgvc-aircraft/#format
• Paper: Towards a Detailed Understanding of Objects and Scenes in Natural Images

Dataset Structure

The dataset contains 10,200 aircraft images, with 100 images per aircraft model, covering 102 different aircraft model variants, most of which are aircraft. Each image's main aircraft has a tight bounding box and hierarchical model labels.

Aircraft models are organized into four levels:

• Model, e.g., Boeing 737-76J. This level is not used for evaluation because some models are visually indistinguishable.
• Variant, e.g., Boeing 737-700. A variant merges visually indistinguishable models into a single category. The dataset contains 102 different variants.
• Family, e.g., Boeing 737. The dataset contains 70 different families.
• Manufacturer, e.g., Boeing. The dataset contains 41 different manufacturers.

The data are split into three equally sized training, validation, and test subsets. The first two subsets can be used for development, while the final subset should be used only for final evaluation.

Dataset Creation

The dataset creation started at the 2012 Johns Hopkins University CLSP summer workshop Towards a Detailed Understanding of Objects and Scenes in Natural Images, with participants including Matthew B. Blaschko, Ross B. Girshick, Juho Kannala, Iasonas Kokkinos, Siddharth Mahendran, Subhransu Maji, Sammy Mohamed, Esa Rahtu, Naomi Saphra, Karen Simonyan, Ben Taskar, Andrea Vedaldi, and David Weiss.

Special thanks to Pekka Rantalankila for help in creating the aircraft hierarchy.

Many thanks to the photographers who generously contributed their images for research purposes. Each photographer's airliners.net page is listed:

• Mick Bajcar
• Aldo Bidini
• Wim Callaert
• Tommy Desmet
• Thomas Posch
• James Richard Covington
• Gerry Stegmeier
• Ben Wang
• Darren Wilson
• Konstantin von Wedelstaedt

Please note that these images are for non‑commercial research purposes only. The original authors retain the copyright of the respective images; for other uses, contact the original authors.

Citation

BibTeX:

bibtex @techreport{maji13fine-grained, title = {Fine-Grained Visual Classification of Aircraft}, author = {S. Maji and J. Kannala and E. Rahtu and M. Blaschko and A. Vedaldi}, year = {2013}, archivePrefix = {arXiv}, eprint = {1306.5151}, primaryClass = "cs-cv", }