This paper proposes a deep neural network (DNN) for piece-wise planar
depthmap reconstruction from a single RGB image. While DNNs have brought
remarkable progress to single-image depth prediction, piece-wise planar
depthmap reconstruction requires a structured geometry representation, and has
been a difficult task to master even for DNNs. The proposed end-to-end DNN
learns to directly infer a set of plane parameters and corresponding plane
segmentation masks from a single RGB image. We have generated more than 50,000
piece-wise planar depthmaps for training and testing from ScanNet, a
large-scale RGBD video database. Our qualitative and quantitative evaluations
demonstrate that the proposed approach outperforms baseline methods in terms of
both plane segmentation and depth estimation accuracy. To the best of our
knowledge, this paper presents the first end-to-end neural architecture for
piece-wise planar reconstruction from a single RGB image. Code and data are
available at https://github.com/art-programmer/PlaneNet.
Description
[1804.06278] PlaneNet: Piece-wise Planar Reconstruction from a Single RGB Image
%0 Generic
%1 liu2018planenet
%A Liu, Chen
%A Yang, Jimei
%A Ceylan, Duygu
%A Yumer, Ersin
%A Furukawa, Yasutaka
%D 2018
%K 3d cnn image neural plane reconstruction single
%T PlaneNet: Piece-wise Planar Reconstruction from a Single RGB Image
%U http://arxiv.org/abs/1804.06278
%X This paper proposes a deep neural network (DNN) for piece-wise planar
depthmap reconstruction from a single RGB image. While DNNs have brought
remarkable progress to single-image depth prediction, piece-wise planar
depthmap reconstruction requires a structured geometry representation, and has
been a difficult task to master even for DNNs. The proposed end-to-end DNN
learns to directly infer a set of plane parameters and corresponding plane
segmentation masks from a single RGB image. We have generated more than 50,000
piece-wise planar depthmaps for training and testing from ScanNet, a
large-scale RGBD video database. Our qualitative and quantitative evaluations
demonstrate that the proposed approach outperforms baseline methods in terms of
both plane segmentation and depth estimation accuracy. To the best of our
knowledge, this paper presents the first end-to-end neural architecture for
piece-wise planar reconstruction from a single RGB image. Code and data are
available at https://github.com/art-programmer/PlaneNet.
@misc{liu2018planenet,
abstract = {This paper proposes a deep neural network (DNN) for piece-wise planar
depthmap reconstruction from a single RGB image. While DNNs have brought
remarkable progress to single-image depth prediction, piece-wise planar
depthmap reconstruction requires a structured geometry representation, and has
been a difficult task to master even for DNNs. The proposed end-to-end DNN
learns to directly infer a set of plane parameters and corresponding plane
segmentation masks from a single RGB image. We have generated more than 50,000
piece-wise planar depthmaps for training and testing from ScanNet, a
large-scale RGBD video database. Our qualitative and quantitative evaluations
demonstrate that the proposed approach outperforms baseline methods in terms of
both plane segmentation and depth estimation accuracy. To the best of our
knowledge, this paper presents the first end-to-end neural architecture for
piece-wise planar reconstruction from a single RGB image. Code and data are
available at https://github.com/art-programmer/PlaneNet.},
added-at = {2019-12-03T16:50:45.000+0100},
author = {Liu, Chen and Yang, Jimei and Ceylan, Duygu and Yumer, Ersin and Furukawa, Yasutaka},
biburl = {https://www.bibsonomy.org/bibtex/2c29aff92b988f9b0c482cf1fc3fc3046/kluger},
description = {[1804.06278] PlaneNet: Piece-wise Planar Reconstruction from a Single RGB Image},
interhash = {91c460396d584d48eb1af68040cdf6e7},
intrahash = {c29aff92b988f9b0c482cf1fc3fc3046},
keywords = {3d cnn image neural plane reconstruction single},
note = {cite arxiv:1804.06278Comment: CVPR 2018},
timestamp = {2019-12-03T16:50:45.000+0100},
title = {PlaneNet: Piece-wise Planar Reconstruction from a Single RGB Image},
url = {http://arxiv.org/abs/1804.06278},
year = 2018
}