A benchmark dataset for deep learning-based airplane detection: HRPlanes [Download][(https://dergipark.org.tr/en/pub/ijeg/article/1107890)]
This repo contains weights of YOLOv4 and Faster R-CNN networks trained with HRPlanes dataset. YOLOv4 training have been performed using Darknet (https://github.com/AlexeyAB/darknet). Faster R-CNN have been trained using TensorFlow Object Detection API v1.13 (https://github.com/tensorflow/models/tree/r1.13.0).
The imagery required for the dataset has been obtained from Google Earth. We have downloaded 4800 x 2703 sized 3092 RGB images from the biggest airports of the world such as Paris-Charles de Gaulle, John F. Kennedy, Frankfurt, Istanbul, Madrid, Dallas, Las Vegas and Amsterdam Airports and aircraft boneyards like Davis-Monthan Air Force Base. Dataset images were annotated manually by creating bounding boxes for each airplane using formerly HyperLabel software which still provides annotation services as Plainsight (https://app.plainsight.ai/). Quality control of each label was conducted by visual inspection of independent analysts who were not included in the labelling procedure. A total of 18,477 airplanes have been labelled. A sample image and corresponding minimum boxes for airplanes can be seen the figure. The dataset has been approximately split as 70% (2166 images), 20% (615 images) and 10% (311 images) for training, validation and testing, respectively.
 |
 |
 |
 |
 |
 |
Use of the Google Earth images must respect the "Google Earth" terms of use. All images and their associated annotations can be used for academic purposes only, but any commercial use is prohibited.
Released under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) License.
If you use the this repository, please cite our paper given below:
Bakirman, T., and Sertel, E. (2023). A benchmark dataset for deep learning-based airplane detection: HRPlanes. International Journal of Engineering and Geosciences, 8(3), 212-223. https://doi.org/10.26833/ijeg.1107890