Multispectral human co-segmentation via joint convolutional neural networks

Sungil Choi; Seungryong Kim; Kihong Park; Kwanghoon Sohn

doi:10.1109/ICIP.2017.8296856

Multispectral human co-segmentation via joint convolutional neural networks

Sungil Choi, Seungryong Kim, Kihong Park, Kwanghoon Sohn

Department of Electrical and Electronic Engineering

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

2 Citations (Scopus)

Abstract

We present a novel human body co-segmentation method for unregistered multispectral, color and thermal, images by leveraging CNNs. The main challenges for that tasks are no-alignment between color and thermal images and an absent of ground truth human segmentation labels. To solve these limitations, our key-insight is to formulate the segmentation network for each modality that solve two sub-tasks, correspondence and classification, in a joint and iterative manner. We formulate the learning framework between multispectral images in a way that training labels for one modality are used to learn the network for the other modality. We estimate dense correspondences between multispectral image pairs using intermediate convolutional activations of CNNs and perform human segmentation for each modality through the conditional random fields (CRF) optimization using unary and pairwise fusion. These two steps are formulated as an iterative framework, enables the network to converge on an optimal solution. Experimental results show that our proposed method outperforms conventional state-of-the-art methods on the VAP benchmark consisting of unregistered multispectral color and thermal images.

Original language	English
Title of host publication	2017 IEEE International Conference on Image Processing, ICIP 2017 - Proceedings
Publisher	IEEE Computer Society
Pages	3115-3119
Number of pages	5
ISBN (Electronic)	9781509021758
DOIs	https://doi.org/10.1109/ICIP.2017.8296856
Publication status	Published - 2018 Feb 20
Event	24th IEEE International Conference on Image Processing, ICIP 2017 - Beijing, China Duration: 2017 Sep 17 → 2017 Sep 20

Publication series

Name	Proceedings - International Conference on Image Processing, ICIP
Volume	2017-September
ISSN (Print)	1522-4880

Other

Other	24th IEEE International Conference on Image Processing, ICIP 2017
Country	China
City	Beijing
Period	17/9/17 → 17/9/20

Bibliographical note

Funding Information:
We proposed the human body co-segmentation method for unregistered multi-spectral color and thermal image by leveraging CNNs. We formulated the human co-segmentation problem as two sub-problems, dense correspondence and segmentation, in a joint and iterative manner. We formulated the learning framework such that training labels for one modality are gradually updated as an iteration progress to learn the other modality network. To realize that task, dense correspondences are estimated using intermediate convolutional activations of segmentation network, and with this flow field, the human body segmentation labels are fused in a unary and pairwise term within CRF optimization. Experimental results demonstrated that our method provides highly accurate segmentation results in comparison to state-of-the-art methods even without pixel-level human annotations. 5. ACKNOWLEDGEMENTS This work was supported by Institute for Information communications Technology Promotion(IITP) grant funded by the Korea government(MSIP)(No.2016-0-00197)

All Science Journal Classification (ASJC) codes

Software
Computer Vision and Pattern Recognition
Signal Processing

Access to Document

10.1109/ICIP.2017.8296856

Fingerprint Dive into the research topics of 'Multispectral human co-segmentation via joint convolutional neural networks'. Together they form a unique fingerprint.

Cite this

Choi, S., Kim, S., Park, K., & Sohn, K. (2018). Multispectral human co-segmentation via joint convolutional neural networks. In 2017 IEEE International Conference on Image Processing, ICIP 2017 - Proceedings (pp. 3115-3119). (Proceedings - International Conference on Image Processing, ICIP; Vol. 2017-September). IEEE Computer Society. https://doi.org/10.1109/ICIP.2017.8296856

@inproceedings{4c6acb582c5049fa8532f83127f7cab4,

title = "Multispectral human co-segmentation via joint convolutional neural networks",

abstract = "We present a novel human body co-segmentation method for unregistered multispectral, color and thermal, images by leveraging CNNs. The main challenges for that tasks are no-alignment between color and thermal images and an absent of ground truth human segmentation labels. To solve these limitations, our key-insight is to formulate the segmentation network for each modality that solve two sub-tasks, correspondence and classification, in a joint and iterative manner. We formulate the learning framework between multispectral images in a way that training labels for one modality are used to learn the network for the other modality. We estimate dense correspondences between multispectral image pairs using intermediate convolutional activations of CNNs and perform human segmentation for each modality through the conditional random fields (CRF) optimization using unary and pairwise fusion. These two steps are formulated as an iterative framework, enables the network to converge on an optimal solution. Experimental results show that our proposed method outperforms conventional state-of-the-art methods on the VAP benchmark consisting of unregistered multispectral color and thermal images.",

author = "Sungil Choi and Seungryong Kim and Kihong Park and Kwanghoon Sohn",

note = "Funding Information: We proposed the human body co-segmentation method for unregistered multi-spectral color and thermal image by leveraging CNNs. We formulated the human co-segmentation problem as two sub-problems, dense correspondence and segmentation, in a joint and iterative manner. We formulated the learning framework such that training labels for one modality are gradually updated as an iteration progress to learn the other modality network. To realize that task, dense correspondences are estimated using intermediate convolutional activations of segmentation network, and with this flow field, the human body segmentation labels are fused in a unary and pairwise term within CRF optimization. Experimental results demonstrated that our method provides highly accurate segmentation results in comparison to state-of-the-art methods even without pixel-level human annotations. 5. ACKNOWLEDGEMENTS This work was supported by Institute for Information communications Technology Promotion(IITP) grant funded by the Korea government(MSIP)(No.2016-0-00197); 24th IEEE International Conference on Image Processing, ICIP 2017 ; Conference date: 17-09-2017 Through 20-09-2017",

year = "2018",

month = feb,

day = "20",

doi = "10.1109/ICIP.2017.8296856",

language = "English",

series = "Proceedings - International Conference on Image Processing, ICIP",

publisher = "IEEE Computer Society",

pages = "3115--3119",

booktitle = "2017 IEEE International Conference on Image Processing, ICIP 2017 - Proceedings",

address = "United States",

}

Choi, S, Kim, S, Park, K & Sohn, K 2018, Multispectral human co-segmentation via joint convolutional neural networks. in 2017 IEEE International Conference on Image Processing, ICIP 2017 - Proceedings. Proceedings - International Conference on Image Processing, ICIP, vol. 2017-September, IEEE Computer Society, pp. 3115-3119, 24th IEEE International Conference on Image Processing, ICIP 2017, Beijing, China, 17/9/17. https://doi.org/10.1109/ICIP.2017.8296856

Multispectral human co-segmentation via joint convolutional neural networks. / Choi, Sungil; Kim, Seungryong; Park, Kihong; Sohn, Kwanghoon.

2017 IEEE International Conference on Image Processing, ICIP 2017 - Proceedings. IEEE Computer Society, 2018. p. 3115-3119 (Proceedings - International Conference on Image Processing, ICIP; Vol. 2017-September).

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

TY - GEN

T1 - Multispectral human co-segmentation via joint convolutional neural networks

AU - Choi, Sungil

AU - Kim, Seungryong

AU - Park, Kihong

AU - Sohn, Kwanghoon

N1 - Funding Information: We proposed the human body co-segmentation method for unregistered multi-spectral color and thermal image by leveraging CNNs. We formulated the human co-segmentation problem as two sub-problems, dense correspondence and segmentation, in a joint and iterative manner. We formulated the learning framework such that training labels for one modality are gradually updated as an iteration progress to learn the other modality network. To realize that task, dense correspondences are estimated using intermediate convolutional activations of segmentation network, and with this flow field, the human body segmentation labels are fused in a unary and pairwise term within CRF optimization. Experimental results demonstrated that our method provides highly accurate segmentation results in comparison to state-of-the-art methods even without pixel-level human annotations. 5. ACKNOWLEDGEMENTS This work was supported by Institute for Information communications Technology Promotion(IITP) grant funded by the Korea government(MSIP)(No.2016-0-00197)

PY - 2018/2/20

Y1 - 2018/2/20

N2 - We present a novel human body co-segmentation method for unregistered multispectral, color and thermal, images by leveraging CNNs. The main challenges for that tasks are no-alignment between color and thermal images and an absent of ground truth human segmentation labels. To solve these limitations, our key-insight is to formulate the segmentation network for each modality that solve two sub-tasks, correspondence and classification, in a joint and iterative manner. We formulate the learning framework between multispectral images in a way that training labels for one modality are used to learn the network for the other modality. We estimate dense correspondences between multispectral image pairs using intermediate convolutional activations of CNNs and perform human segmentation for each modality through the conditional random fields (CRF) optimization using unary and pairwise fusion. These two steps are formulated as an iterative framework, enables the network to converge on an optimal solution. Experimental results show that our proposed method outperforms conventional state-of-the-art methods on the VAP benchmark consisting of unregistered multispectral color and thermal images.

AB - We present a novel human body co-segmentation method for unregistered multispectral, color and thermal, images by leveraging CNNs. The main challenges for that tasks are no-alignment between color and thermal images and an absent of ground truth human segmentation labels. To solve these limitations, our key-insight is to formulate the segmentation network for each modality that solve two sub-tasks, correspondence and classification, in a joint and iterative manner. We formulate the learning framework between multispectral images in a way that training labels for one modality are used to learn the network for the other modality. We estimate dense correspondences between multispectral image pairs using intermediate convolutional activations of CNNs and perform human segmentation for each modality through the conditional random fields (CRF) optimization using unary and pairwise fusion. These two steps are formulated as an iterative framework, enables the network to converge on an optimal solution. Experimental results show that our proposed method outperforms conventional state-of-the-art methods on the VAP benchmark consisting of unregistered multispectral color and thermal images.

UR - http://www.scopus.com/inward/record.url?scp=85045326472&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85045326472&partnerID=8YFLogxK

U2 - 10.1109/ICIP.2017.8296856

DO - 10.1109/ICIP.2017.8296856

M3 - Conference contribution

AN - SCOPUS:85045326472

T3 - Proceedings - International Conference on Image Processing, ICIP

SP - 3115

EP - 3119

BT - 2017 IEEE International Conference on Image Processing, ICIP 2017 - Proceedings

PB - IEEE Computer Society

T2 - 24th IEEE International Conference on Image Processing, ICIP 2017

Y2 - 17 September 2017 through 20 September 2017

ER -