kornia/orinet

Pretrained weights for OriNet, used by kornia.feature.LAFOrienter.

OriNet is a CNN that regresses the dominant gradient orientation for a keypoint patch, trained with a Siamese setup to produce consistent orientations across viewpoints. Companion to AffNet. ECCV 2018.

Original repo: ducha-aiki/affnet

Weights

File	Description
OriNet.pth	Default checkpoint

Citation

@inproceedings{AffNet2018,
    author    = {D. Mishkin and F. Radenovic and J. Matas},
    title     = {{Repeatability is Not Enough}: Learning Affine Regions
                 via Discriminability},
    booktitle = {ECCV},
    year      = {2018}
}