Calibrated and Partially Calibrated Semi-Generalized Homographies

Bhayani, Snehal; Sattler, Torsten; Baráth, Dániel; Beliansky, Patrik; Heikkilä, Janne; Kúkelová, Zuzana

doi:10.1109/iccv48922.2021.00588

Cited by 8 publications

(9 citation statements)

References 57 publications

(82 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the second configuration, we have three 2D-2D point correspondences p j ↔ (q ij , t Gi ) and two 2D-3D point correspondences p j ↔ X G j . Thus, there are three possible generalized camera configurations, i.e., [1], [2] and [3]. For all three camera configurations, we obtained the smallest solvers using the Rotation & translation parameterization.…”

Section: H32fmentioning

confidence: 99%

“…Next, we review the coordinate transformation and the parameterization proposed in [13] and then describe our proposed solver. We obtain the same solver for all three configurations [1], [2] and [3].…”

Section: H32fmentioning

confidence: 99%

“…The proposed E 4+2 and Ef 5+2 solvers were impractical for real-time applications since they perform eliminations of huge matrices and have running times of 1.2ms and 13.6ms, respectively. Recently, [2] showed that for planar scenes, the semi-generalized relative pose problem can be, after using suitable parameterizations and after eliminating some unknowns, solved efficiently by finding the roots of a single variable polynomial. The authors proposed such efficient minimal solutions for calibrated and partially-calibrated semi-generalized homography estimation and all different configurations of 2D-2D correspondences in a generalized camera.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Partially calibrated semi-generalized pose from hybrid point correspondences

Bhayani

Sattler

Larsson

et al. 2023

2023 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)

Self Cite

View full text Add to dashboard Cite

In this paper we study the problem of estimating the semi-generalized pose of a partially calibrated camera, i.e., the pose of a perspective camera with unknown focal length w.r.t. a generalized camera, from a hybrid set of 2D-2D and 2D-3D point correspondences. We study all possible camera configurations within the generalized camera system. To derive practical solvers to previously unsolved challenging configurations, we test different parameterizations as well as different solving strategies based on state-of-theart methods for generating efficient polynomial solvers. We evaluate the three most promising solvers, i.e., the H51f solver with five 2D-2D correspondences and one 2D-3D correspondence viewed by the same camera inside generalized camera, the H32f solver with three 2D-2D and two 2D-3D correspondences, and the H13f solver with one 2D-2D and three 2D-3D correspondences, on synthetic and real data. We show that in the presence of noise in the 3D points these solvers provide better estimates than the corresponding absolute pose solvers.

show abstract

Section: H32fmentioning

confidence: 99%

Section: H32fmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Partially calibrated semi-generalized pose from hybrid point correspondences

Bhayani

Sattler

Larsson

et al. 2023

2023 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)

Self Cite

View full text Add to dashboard Cite

show abstract

“…We considered the two cameras as a generalized one and estimated the pose between this camera and the left image of the next frame. We used the generalized essential matrix [69] (E4+2) and homography [8] (H3+2) solvers. For finding a single E4+2 or H3+2, we used GC-RANSAC [3].…”

Section: Pose From Generalized Cameramentioning

confidence: 99%

“…and med. rotation R (degrees) and position p (meters) errors on 23 190 image pairs from the KITTI dataset obtained by generalized essential matrix (E4+2)[69] and homography estimation[8] (H3+2); and by selecting the pose obtained from a generalized E matrix and a set of generalized homographies (E + mult. Hs).…”

mentioning

confidence: 99%

Progressive-X+: Clustering in the Consensus Space

Baráth¹,

Rozumny²,

Eichhardt³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

We propose Progressive-X + , 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 pointto-model assignments. Dominant instances are found via RANSAC-like sampling and a consolidation process driven by a model quality function considering previously proposed instances. New ones are found by clustering in the consensus space. This new formulation leads to a simple iterative algorithm with state-of-the-art accuracy while running in real-time on a number of vision problems. Also, we propose a sampler reflecting the fact that real-world data tend to form spatially coherent structures. The sampler returns connected components in a progressively growing neighborhood-graph. We present a number of applications where the use of multiple geometric models improves accuracy. These include using multiple homographies to estimate relative poses for global SfM; pose estimation from generalized homographies; and trajectory estimation of fast-moving objects.

show abstract

ViewFormer: NeRF-Free Neural Rendering from Few Images Using Transformers

Kulhánek

Derner

Sattler

et al. 2022

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Novel view synthesis is a long-standing problem. In this work, we consider a variant of the problem where we are given only a few context views sparsely covering a scene or an object. The goal is to predict novel viewpoints in the scene, which requires learning priors. The current state of the art is based on Neural Radiance Field (NeRF), and while achieving impressive results, the methods suffer from long training times as they require evaluating millions of 3D point samples via a neural network for each image. We propose a 2D-only method that maps multiple context views and a query pose to a new image in a single pass of a neural network. Our model uses a two-stage architecture consisting of a codebook and a transformer model. The codebook is used to embed individual images into a smaller latent space, and the transformer solves the view synthesis task in this more compact space. To train our model efficiently, we introduce a novel branching attention mechanism that allows us to use the same model not only for neural rendering but also for camera pose estimation. Experimental results on real-world scenes show that our approach is competitive compared to NeRF-based methods while not reasoning explicitly in 3D, and it is faster to train.

show abstract

Calibrated and Partially Calibrated Semi-Generalized Homographies

Cited by 8 publications

References 57 publications

Partially calibrated semi-generalized pose from hybrid point correspondences

Partially calibrated semi-generalized pose from hybrid point correspondences

Progressive-X+: Clustering in the Consensus Space

ViewFormer: NeRF-Free Neural Rendering from Few Images Using Transformers

Contact Info

Product

Resources

About