Lifelong Graph Learning (CVPR 2022) [Feature Matching]
BSD-3-CLAUSE License
This repo contains the source code for the feature matching application (Sec. 7) in "Lifelong Graph Learning." Chen Wang, Yuheng Qiu, Dasong Gao, Sebastian Scherer. CVPR 2022.
The TartanAir dataset is required for both training and testing. The dataset should be aranged as follows:
$DATASET_ROOT/
tartanair/
abandonedfactory_night/
...
Training and evaluates the method with default setting:
$ python train.py --data-root <DATASET_ROOT> --method <FGN/GAT>
--method
option is used to switch between FGN-based (ours) and GAT-based (SuperGlue) graph matcher--eval-freq
). Results will be logged to the console.--log-dir
is specified, TensorBoard will be activated to show visualization and evaluation results instead (under "TEXT" tab).$ python train.py -h
.@inproceedings{wang2022lifelong,
title={Lifelong graph learning},
author={Wang, Chen and Qiu, Yuheng and Gao, Dasong and Scherer, Sebastian},
booktitle={2022 Conference on Computer Vision and Pattern Recognition (CVPR)},
year={2022}
}