kornia/keynet

Pretrained weights for Key.Net,

used by kornia.feature.KeyNet.

Key.Net is a keypoint detector that combines handcrafted (Difference of Gaussians) filters

with learned filters inside a shallow multi-scale CNN, trained to maximise repeatability

under homographic transformations. ICCV 2019.

Original repo: axelBarroso/Key.Net-Pytorch

Weights

| File | Description |

|------|-------------|

| keynet_pytorch.pth | Default checkpoint |

Citation

@inproceedings{KeyNet2019,
    author    = {Barroso-Laguna, Axel and Riba, Edgar
                 and Ponsa, Daniel and Mikolajczyk, Krystian},
    title     = {{Key.Net}: Keypoint Detection by Handcrafted and Learned CNN Filters},
    booktitle = {ICCV},
    year      = {2019}
}