Deep Aerial Image Matching Implementation

This is the official implementation of the paper:

J.-H. Park, W.-J Nam and S.-W Lee, "A Two-Stream Symmetric Network with Bidirectional Ensemble for Aerial Image Matching," Remote Sens., 2020, Vol. 12, No. 6, pp. 465
[Journal][arXiv]

Required package

• Python 3
• PyTorch, torchvision
• pretrainedmodels
• scikit-image, pandas, opencv
• termcolor, tqdm
• googledrivedownloader

Getting started

• demo.py demonstrates the results on the samples of aerial image dataset
• train.py is the main training script
• eval_pck.py evaluates on the aerial image dataset

Trained models

Note that, models must be downloaded to the 'trained_models' folder.

Paper Citation

@misc{park2020aerial,
title={A Two-Stream Symmetric Network with Bidirectional Ensemble for Aerial Image Matching},
author={Jae-Hyun Park and Woo-Jeoung Nam and Seong-Whan Lee},
year={2020},
eprint={2002.01325},
archivePrefix={arXiv},
primaryClass={cs.CV}
}