This is a PyTorch implementation of our paper:

"HyNet: Learning Local Descriptor with Hybrid Similarity Measure and Triplet Loss"
Yurun Tian, Axel Barroso-Laguna, Tony Ng, Vassileios Balntas, Krystian Mikolajczyk. NeurIPS 2020. [arXiv]

• Python 3
• PyTorch tested on 1.4.0-1.9.0
• opencv-python (cv2) tested on 3.3.0.10
• numpy
• PIL
• tqdm

• We provide codes for training on the UBC data set and the HPatches data set. The downloaded data should be organised as the following folder structure:

data_root

-- liberty

-- notredame

-- yosemite

-- hpatches-benchmark-master

Specify the training data path and saving path for the code:

python train.py --data_root=data_root --network_root=save_root

• To accelerate the training, all the data needed will be generated and saved at the first run.

If you use this repository in your work, please cite our paper:

@inproceedings{hynet2020,
 author = {Tian, Yurun and Barroso Laguna, Axel and Ng, Tony and Balntas, Vassileios and Mikolajczyk, Krystian},
 title = {HyNet: Learning Local Descriptor with Hybrid Similarity Measure and Triplet Loss},
 booktitle = {NeurIPS},
 year      = {2020}
}