kornia/aliked

Pretrained weights for ALIKED (A Lighter Keypoint and Descriptor Extraction Network via Deformable Transformation), used by kornia.feature.ALIKED.

ALIKED uses deformable convolutional heads to detect keypoints that are inherently invariant to affine deformations, with a sparse descriptor extraction step. IEEE Transactions on Instrumentation and Measurement, 2023.

Original repo: Shiaoming/ALIKED Original license: BSD-3-Clause

Weights

File Variant
aliked-t16.pth Tiny, 16-dim
aliked-n16.pth Normal, 16-dim
aliked-n16rot.pth Normal, 16-dim, rotation-robust
aliked-n32.pth Normal, 32-dim

Citation

@article{zhao2023aliked,
    title   = {{ALIKED}: A Lighter Keypoint and Descriptor Extraction Network
               via Deformable Transformation},
    author  = {Xiaoming Zhao and Xingming Wu and Weihai Chen and Peter C.Y. Chen
               and Qingsong Xu and Zhengguo Li},
    journal = {IEEE Transactions on Instrumentation and Measurement},
    year    = {2023}
}
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support