Augmenting depth estimation from deep convolutional neural network using multi-spectral photometric stereo

Yisong Luo; Hengchao Jiao; Lin Qi; Junyu Dong; Shu Zhang; Hui Yu

doi:10.1109/UIC-ATC.2017.8397464

Augmenting depth estimation from deep convolutional neural network using multi-spectral photometric stereo

Yisong Luo, Hengchao Jiao, Lin Qi, Junyu Dong, Shu Zhang, Hui Yu

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

315 Downloads (Pure)

Abstract

Multi-Spectral Photometric Stereo can recover surface normals from a single image, but requires an initial estimate of the normals due to the tangle of the illumination, reflectance and camera responses on each of the RGB channels. Instead of employing a depth sensor or binocular stereo device, in this paper, we propose a method to estimate fine-scale geometry structures with the popular Deep Convolutional Neural Networks (CNNs). We train the network with rendered images of synthetic 3D objects, and apply the trained model with real world data. The CNN is used to estimate a rough prediction of the depth, then the normals from Multi-Spectral Photometric Stereo are progressively refined accordingly. Experiments demonstrate the competitive results of our method for improving the depth estimation.

Original language	English
Title of host publication	2017 IEEE SmartWorld, Ubiquitous Intelligence & Computing, Advanced & Trusted Computed, Scalable Computing & Communications, Cloud & Big Data Computing, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI)
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	978-1-5386-0435-9, 978-1-5386-0434-2
ISBN (Print)	978-1-5386-1591-1
DOIs	https://doi.org/10.1109/UIC-ATC.2017.8397464
Publication status	Published - 28 Jun 2018
Event	14th IEEE International Conference on Ubiquitous Intelligence and Computing - San Francisco, United States Duration: 4 Aug 2017 → 8 Aug 2017 http://ieee-smartworld.org/2017/uic/

Conference

Conference	14th IEEE International Conference on Ubiquitous Intelligence and Computing
Abbreviated title	UIC 2017
Country/Territory	United States
City	San Francisco
Period	4/08/17 → 8/08/17
Internet address	http://ieee-smartworld.org/2017/uic/

Access to Document

10.1109/UIC-ATC.2017.8397464

Augmenting Depth Estimation from Deep Convolutional Neural Network_UIC2017
© © 2017 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.
Accepted author manuscript (Post-print), 1.24 MB

Cite this

Luo, Y., Jiao, H., Qi, L., Dong, J., Zhang, S., & Yu, H. (2018). Augmenting depth estimation from deep convolutional neural network using multi-spectral photometric stereo. In 2017 IEEE SmartWorld, Ubiquitous Intelligence & Computing, Advanced & Trusted Computed, Scalable Computing & Communications, Cloud & Big Data Computing, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI) Article 17878684 Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/UIC-ATC.2017.8397464

Luo, Yisong ; Jiao, Hengchao ; Qi, Lin et al. / Augmenting depth estimation from deep convolutional neural network using multi-spectral photometric stereo. 2017 IEEE SmartWorld, Ubiquitous Intelligence & Computing, Advanced & Trusted Computed, Scalable Computing & Communications, Cloud & Big Data Computing, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI). Institute of Electrical and Electronics Engineers Inc., 2018.

@inproceedings{fe8aa6e0ceef4fb1bf0df2f95e8381f1,

title = "Augmenting depth estimation from deep convolutional neural network using multi-spectral photometric stereo",

abstract = "Multi-Spectral Photometric Stereo can recover surface normals from a single image, but requires an initial estimate of the normals due to the tangle of the illumination, reflectance and camera responses on each of the RGB channels. Instead of employing a depth sensor or binocular stereo device, in this paper, we propose a method to estimate fine-scale geometry structures with the popular Deep Convolutional Neural Networks (CNNs). We train the network with rendered images of synthetic 3D objects, and apply the trained model with real world data. The CNN is used to estimate a rough prediction of the depth, then the normals from Multi-Spectral Photometric Stereo are progressively refined accordingly. Experiments demonstrate the competitive results of our method for improving the depth estimation.",

author = "Yisong Luo and Hengchao Jiao and Lin Qi and Junyu Dong and Shu Zhang and Hui Yu",

year = "2018",

month = jun,

day = "28",

doi = "10.1109/UIC-ATC.2017.8397464",

language = "English",

isbn = "978-1-5386-1591-1",

booktitle = "2017 IEEE SmartWorld, Ubiquitous Intelligence & Computing, Advanced & Trusted Computed, Scalable Computing & Communications, Cloud & Big Data Computing, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI)",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

address = "United States",

note = "14th IEEE International Conference on Ubiquitous Intelligence and Computing, UIC 2017 ; Conference date: 04-08-2017 Through 08-08-2017",

url = "http://ieee-smartworld.org/2017/uic/",

}

Luo, Y, Jiao, H, Qi, L, Dong, J, Zhang, S & Yu, H 2018, Augmenting depth estimation from deep convolutional neural network using multi-spectral photometric stereo. in 2017 IEEE SmartWorld, Ubiquitous Intelligence & Computing, Advanced & Trusted Computed, Scalable Computing & Communications, Cloud & Big Data Computing, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI)., 17878684, Institute of Electrical and Electronics Engineers Inc., 14th IEEE International Conference on Ubiquitous Intelligence and Computing, San Francisco, California, United States, 4/08/17. https://doi.org/10.1109/UIC-ATC.2017.8397464

Augmenting depth estimation from deep convolutional neural network using multi-spectral photometric stereo. / Luo, Yisong; Jiao, Hengchao; Qi, Lin et al.
2017 IEEE SmartWorld, Ubiquitous Intelligence & Computing, Advanced & Trusted Computed, Scalable Computing & Communications, Cloud & Big Data Computing, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI). Institute of Electrical and Electronics Engineers Inc., 2018. 17878684.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Augmenting depth estimation from deep convolutional neural network using multi-spectral photometric stereo

AU - Luo, Yisong

AU - Jiao, Hengchao

AU - Qi, Lin

AU - Dong, Junyu

AU - Zhang, Shu

AU - Yu, Hui

PY - 2018/6/28

Y1 - 2018/6/28

N2 - Multi-Spectral Photometric Stereo can recover surface normals from a single image, but requires an initial estimate of the normals due to the tangle of the illumination, reflectance and camera responses on each of the RGB channels. Instead of employing a depth sensor or binocular stereo device, in this paper, we propose a method to estimate fine-scale geometry structures with the popular Deep Convolutional Neural Networks (CNNs). We train the network with rendered images of synthetic 3D objects, and apply the trained model with real world data. The CNN is used to estimate a rough prediction of the depth, then the normals from Multi-Spectral Photometric Stereo are progressively refined accordingly. Experiments demonstrate the competitive results of our method for improving the depth estimation.

AB - Multi-Spectral Photometric Stereo can recover surface normals from a single image, but requires an initial estimate of the normals due to the tangle of the illumination, reflectance and camera responses on each of the RGB channels. Instead of employing a depth sensor or binocular stereo device, in this paper, we propose a method to estimate fine-scale geometry structures with the popular Deep Convolutional Neural Networks (CNNs). We train the network with rendered images of synthetic 3D objects, and apply the trained model with real world data. The CNN is used to estimate a rough prediction of the depth, then the normals from Multi-Spectral Photometric Stereo are progressively refined accordingly. Experiments demonstrate the competitive results of our method for improving the depth estimation.

UR - http://smart-city-sjsu.net//conference/docs/UIC-Conference-Program-Final.htm

U2 - 10.1109/UIC-ATC.2017.8397464

DO - 10.1109/UIC-ATC.2017.8397464

M3 - Conference contribution

SN - 978-1-5386-1591-1

BT - 2017 IEEE SmartWorld, Ubiquitous Intelligence & Computing, Advanced & Trusted Computed, Scalable Computing & Communications, Cloud & Big Data Computing, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI)

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 14th IEEE International Conference on Ubiquitous Intelligence and Computing

Y2 - 4 August 2017 through 8 August 2017

ER -

Luo Y, Jiao H, Qi L, Dong J, Zhang S, Yu H. Augmenting depth estimation from deep convolutional neural network using multi-spectral photometric stereo. In 2017 IEEE SmartWorld, Ubiquitous Intelligence & Computing, Advanced & Trusted Computed, Scalable Computing & Communications, Cloud & Big Data Computing, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI). Institute of Electrical and Electronics Engineers Inc. 2018. 17878684 doi: 10.1109/UIC-ATC.2017.8397464

Augmenting depth estimation from deep convolutional neural network using multi-spectral photometric stereo

Abstract

Conference

Access to Document

Fingerprint

Cite this