ALIKED: A Lighter Keypoint and Descriptor Extraction Network via Deformable Transformation

Zhao, Xiaoming; Wu, Xingming; Chen, Weihai; Chen, Peter C. Y.; Xu, Qingsong; Li, Zhengguo

doi:10.1109/tim.2023.3271000

Cited by 5 publications

(1 citation statement)

References 49 publications

(155 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…DISK [18] uses reinforcement learning to learn local features end-toend through policy gradient. ALIKE [31] proposed a partially differentiable keypoint extraction module, on which the ALIKED [32] method introduces deformable convolution. Olivia et al [33] introduced a spatial attention mechanism and a distinctiveness score.…”

Section: End-to-end Learned Detectors and Descriptorsmentioning

confidence: 99%

Epipolar Constraint Guided Differentiable Keypoint Detection and Description

Li,

Feng,

et al. 2024

Preprint

View full text Add to dashboard Cite

Sparse local feature matching methods have made significant strides in a variety of visual geometric tasks. Recently, the weakly supervised under epipolar constraint approaches achieve stronger performance compared to the fully supervised methods through the decoupled training describe-then-detect. However, such methods base on the policy gradient to train the detector, ignoring the reliability of the keypoints. Meanwhile, many of the sparse local feature matching methods place more emphasis on accuracy over speed, being unfriendly for real-time applications. To address these issues, we introduce the differentiable keypoint extraction and the dispersity peak loss to generate clean score maps and enhance the reliability of the keypoints in the weakly supervised method. Additionally, we propose a straightforward yet efficient model that achieves a good balance between accuracy and speed. We conduct experiments on multiple public benchmark datasets, achieving higher performance than existing methods. The code will be available at https://github.com/FYL0123/WSDK.

show abstract

Section: End-to-end Learned Detectors and Descriptorsmentioning

confidence: 99%

Epipolar Constraint Guided Differentiable Keypoint Detection and Description

Li,

Feng,

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

Canonical Plane Segmentation Without Annotating Pixel-Level Object Regions for Image Registration

Yoneda,

Irie,

Nishiyama

2024

IEEE Access

View full text Add to dashboard Cite

image registration is a natural choice for simultaneous object pose estimation and object recognition. However, it was not designed to perform object segmentation, which is critical for object-picking applications in warehouse automation scenarios. In this study, we propose a unified 2D image registration framework that simultaneously performs image registration and object segmentation by introducing a deep segmentation network module to the 2D image registration framework. Our method is designed to automatically generate annotations for training the segmentation network module through the process of 2D image registration, that is, no additional manual annotations are required. Specifically, given initial object regions from the 2D image registration results, our method trains the segmentation network module to refine a pseudo-pixel-level object region and remove background pixels based on the pixel-level similarity of an aligned image pair. The experimental results on several picking object datasets demonstrated that the segmentation accuracy of our method was superior to that of existing weakly supervised segmentation methods, and our method simultaneously achieved better performance for object recognition and pose estimation. Furthermore, our segmentation network module smoothly cooperated with many existing 2D image registration techniques.INDEX TERMS Image registration, pseudo-pixel-level object region, weakly supervised segmentation.

show abstract

Self-Supervised Camera Relocalization With Hierarchical Fern Encoding

Lu,

Zhu,

et al. 2024

IEEE Trans. Instrum. Meas.

View full text Add to dashboard Cite

ALIKED: A Lighter Keypoint and Descriptor Extraction Network via Deformable Transformation

Cited by 5 publications

References 49 publications

Epipolar Constraint Guided Differentiable Keypoint Detection and Description

Epipolar Constraint Guided Differentiable Keypoint Detection and Description

Canonical Plane Segmentation Without Annotating Pixel-Level Object Regions for Image Registration

Self-Supervised Camera Relocalization With Hierarchical Fern Encoding

Contact Info

Product

Resources

About