ACuOoOoO / SDGMNet

Codes and models - "SDGMNet: Statistic-based Dynamic Gradient Modulation for Local Descriptor Learning"

SDGMNet

SDGMNet models in Pytorch for paper "SDGMNet: Statistic-based Dynamic Gradient Modulation for Local Descriptor Learning".

Training Code

python train.py --datadir='./UBC'

Results on Image Matching Challenge

Pre-trained models

Pre-trained models and loggers on UBC PhotoTour can be found in pretrained. State-of-the-art models trained on Liberty for Image matching challenge can be found in pretrained/models/PHM0.2.

Losses

We implement several state-of-the-art losses including SDGMNet, SOSNet and HyNet.

Citation

Please cite us if you use these models

@article{ma2021sdgmnet,
  title={SDGMNet: Statistic-based Dynamic Gradient Modulation for Local Descriptor Learning},
  author={Ma, Jiayi and Deng, Yuxin},
  journal={arXiv preprint arXiv:2106.04434},
  year={2021}
}