RAL-Net

This code is the implementation of our ACCV2018 paper Robust Angular Local Descriptor Learning

Enviromental built:

Please install python 3.6, pytorch 0.4.1, opencv 3.4

pip install python==3.6

conda install pytorch=0.4.1 cuda90 -c pytorch

To replicate the result of this paper:

python RAL_Net.py --data-root='your data root' --epochs=10 --batch-size=512 --n-pairs=5000000 --loss-type=RAL_loss --lr=10 --augmentation=True

Result:

Performance comparison on Brown dataset, lower score and perform better

Performance comparison on Hpatches dataset, Higher score and perform better

Performance comparison on Wxbs dataset, Higher score and perform better