English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Matching cross-modality features between images and point clouds is afundamental problem for image-to-point cloud registration. However, due to themodality difference between images and points, it is difficult to learn robustand discriminative cross-modality features by existing metric learning methodsfor feature matching. Instead of applying metric learning on cross-modalitydata, we propose to unify the modality between images and point clouds bypretrained large-scale models first, and then establish robust correspondencewithin the same modality. We show that the intermediate features, calleddiffusion features, extracted by depth-to-image diffusion models aresemantically consistent between images and point clouds, which enables thebuilding of coarse but robust cross-modality correspondences. We furtherextract geometric features on depth maps produced by the monocular depthestimator. By matching such geometric features, we significantly improve theaccuracy of the coarse correspondences produced by diffusion features.Extensive experiments demonstrate that without any task-specific training,direct utilization of both features produces accurate image-to-point cloudregistration. On three public indoor and outdoor benchmarks, the proposedmethod averagely achieves a 20.6 percent improvement in Inlier Ratio, athree-fold higher Inlier Number, and a 48.6 percent improvement in RegistrationRecall than existing state-of-the-arts.

FreeReg: Image-to-Point Cloud Registration Leveraging Pretrained Diffusion Models and Monocular Depth Estimators