Výsledky vyhledávání - "Eichhardt, Ivan"

Report

Finding Geometric Models by Clustering in the Consensus Space

Autor: Barath, Daniel, Rozumny, Denys, Eichhardt, Ivan, Hajder, Levente, Matas, Jiri

We propose a new algorithm for finding an unknown number of geometric models, e.g., homographies. The problem is formalized as finding dominant model instances progressively without forming crisp point-to-model assignments. Dominant instances are fou

Externí odkaz: http://arxiv.org/abs/2103.13875

Zobrazit plný text záznamu

Report

Efficient Initial Pose-graph Generation for Global SfM

Autor: Barath, Daniel, Mishkin, Dmytro, Eichhardt, Ivan, Shipachev, Ilia, Matas, Jiri

We propose ways to speed up the initial pose-graph generation for global Structure-from-Motion algorithms. To avoid forming tentative point correspondences by FLANN and geometric verification by RANSAC, which are the most time-consuming steps of the

Externí odkaz: http://arxiv.org/abs/2011.11986

Zobrazit plný text záznamu

Report

Relative Pose from Deep Learned Depth and a Single Affine Correspondence

Autor: Eichhardt, Ivan, Barath, Daniel

We propose a new approach for combining deep-learned non-metric monocular depth with affine correspondences (ACs) to estimate the relative pose of two calibrated cameras from a single correspondence. Considering the depth information and affine featu

Externí odkaz: http://arxiv.org/abs/2007.10082

Zobrazit plný text záznamu

Report

Optimal Multi-view Correction of Local Affine Frames

Autor: Eichhardt, Ivan, Barath, Daniel

The technique requires the epipolar geometry to be pre-estimated between each image pair. It exploits the constraints which the camera movement implies, in order to apply a closed-form correction to the parameters of the input affinities. Also, it is

Externí odkaz: http://arxiv.org/abs/1905.00519

Zobrazit plný text záznamu