Global, Dense Multiscale Reconstruction for a Billion Points

Ummenhofer, Benjamin; Brox, Thomas

doi:10.1007/s11263-017-1017-7

Cited by 44 publications

(64 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…DeMoN proposes a collection of datasets to train and evaluate deep networks. The dataset contains images from multiple sources, such as RGB-D cameras [7], [24], multiview SfM results [1], [2], [25], [26], and synthetic images [3]. In total, the DeMoN dataset contains 57k image pairs for training and 354 pairs for testing.…”

Section: A Demon Datasetmentioning

confidence: 99%

Flow-Motion and Depth Network for Monocular Stereo and Beyond

Wang

Shen

2020

IEEE Robot. Autom. Lett.

View full text Add to dashboard Cite

We propose a learning-based method 1 that solves monocular stereo and can be extended to fuse depth information from multiple target frames. Given two unconstrained images from a monocular camera with known intrinsic calibration, our network estimates relative camera poses and the depth map of the source image. The core contribution of the proposed method is threefold. First, a network is tailored for static scenes that jointly estimates the optical flow and camera motion. By the joint estimation, the optical flow search space is gradually reduced resulting in an efficient and accurate flow estimation. Second, a novel triangulation layer is proposed to encode the estimated optical flow and camera motion while avoiding common numerical issues caused by epipolar. Third, beyond two-view depth estimation, we further extend the above networks to fuse depth information from multiple target images and estimate the depth map of the source image. To further benefit the research community, we introduce tools to generate photorealistic structure-from-motion datasets such that deep networks can be well trained and evaluated. The proposed method is compared with previous methods and achieves stateof-the-art results within less time. Images from real-world applications and Google Earth are used to demonstrate the generalization ability of the method.

show abstract

Section: A Demon Datasetmentioning

confidence: 99%

Flow-Motion and Depth Network for Monocular Stereo and Beyond

Wang

Shen

2020

IEEE Robot. Autom. Lett.

View full text Add to dashboard Cite

show abstract

“…Scale space has been used to analyze structures in images (e.g., [13,50,29,44]). This has had wide ranging applications in stereo and optical flow [31], reconstruction [20,49], key-point detection in wide-baseline matching [30], design of descriptors for matching [17], shape matching [7], and curve evolution [43], among others.…”

Section: Related Workmentioning

confidence: 99%

Coarse-to-Fine Segmentation with Shape-Tailored Continuum Scale Spaces

Khan

Hong

Yezzi

et al. 2017

2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

View full text Add to dashboard Cite

“…Among off-line methods that globally optimise a batch of images recent examples include work by Fuhrmann and Goesele [8,9] and Ummenhofer and Brox [30]. These approaches have shown remarkable results but are highly prohibitive for real-time applications where processing should be fast and the reconstruction should be updated incrementally.…”

Section: Related Work 21 Multi-scale Reconstructionmentioning

confidence: 99%

Monocular, Real-Time Surface Reconstruction Using Dynamic Level of Detail

Zienkiewicz

Tsiotsios

Davison

et al. 2016

2016 Fourth International Conference on 3D Vision (3DV)

View full text Add to dashboard Cite

We present a scalable, real-time capable method for robust surface reconstruction that explicitly handles multiple scales. As a monocular camera browses a scene, our algorithm processes images as they arrive and incrementally builds a detailed surface model. While most of the existing reconstruction approaches rely on volumetric or point-cloud representations of the environment, we perform depth-map and colour fusion directly into a multi-resolution triangular mesh that can be adaptively tessellated using the concept of Dynamic Level of Detail. Our method relies on least-squares optimisation, which enables a probabilistically sound and principled formulation of the fusion algorithm. We demonstrate that our method is capable of obtaining high quality, close-up reconstruction, as well as capturing overall scene geometry, while being memory and computationally efficient.

show abstract

Global, Dense Multiscale Reconstruction for a Billion Points

Cited by 44 publications

References 31 publications

Flow-Motion and Depth Network for Monocular Stereo and Beyond

Flow-Motion and Depth Network for Monocular Stereo and Beyond

Coarse-to-Fine Segmentation with Shape-Tailored Continuum Scale Spaces

Monocular, Real-Time Surface Reconstruction Using Dynamic Level of Detail

Contact Info

Product

Resources

About