This code is developed based on Caffe
This code is the implementation for the network with the context-based feature reweighting in the paper:
Hyo Jin Kim, Enrique Dunn, and Jan-Michael Frahm. "Learned Contextual Feature Reweighting for Image Geo-Localization". Proc. of IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017. [pdf] [project page]
If you use our codes or models in your research, please cite:
@inproceedings{kim2017crn,
title={Learned Contextual Feature Reweighting for Image Geo-Localization},
author={Kim, Hyo Jin and Dunn, Enrique and Frahm, Jan-Michael},
booktitle={CVPR},
year= {2017}
}
- provide documentation
- provide crn modules for torch
- Learning rate scheduling: Learning rate scheduling is done through babysitting. Whenever the training loss reached a plateau, learning rate was reduced by gamma (as specified in the solver.prototxt).