TUTCRIS - Tampereen teknillinen yliopisto

TUTCRIS

Portrait instance segmentation for mobile devices

Tutkimustuotosvertaisarvioitu

Standard

Portrait instance segmentation for mobile devices. / Zhu, Lingyu; Wang, Tinghuai; Aksu, Emre; Kämäräinen, Joni-Kristian.

2019 IEEE International Conference on Multimedia and Expo, ICME 2019. IEEE, 2019. s. 1630-1635.

Tutkimustuotosvertaisarvioitu

Harvard

Zhu, L, Wang, T, Aksu, E & Kämäräinen, J-K 2019, Portrait instance segmentation for mobile devices. julkaisussa 2019 IEEE International Conference on Multimedia and Expo, ICME 2019. IEEE, Sivut 1630-1635, Shanghai, Kiina, 8/07/19. https://doi.org/10.1109/ICME.2019.00281

APA

Zhu, L., Wang, T., Aksu, E., & Kämäräinen, J-K. (2019). Portrait instance segmentation for mobile devices. teoksessa 2019 IEEE International Conference on Multimedia and Expo, ICME 2019 (Sivut 1630-1635). IEEE. https://doi.org/10.1109/ICME.2019.00281

Vancouver

Zhu L, Wang T, Aksu E, Kämäräinen J-K. Portrait instance segmentation for mobile devices. julkaisussa 2019 IEEE International Conference on Multimedia and Expo, ICME 2019. IEEE. 2019. s. 1630-1635 https://doi.org/10.1109/ICME.2019.00281

Author

Zhu, Lingyu ; Wang, Tinghuai ; Aksu, Emre ; Kämäräinen, Joni-Kristian. / Portrait instance segmentation for mobile devices. 2019 IEEE International Conference on Multimedia and Expo, ICME 2019. IEEE, 2019. Sivut 1630-1635

Bibtex - Lataa

@inproceedings{e79f305fd609429f8c69e9928b4882a3,
title = "Portrait instance segmentation for mobile devices",
abstract = "Accurate and efficient portrait instance segmentation has become a crucial enabler for many multimedia applications on mobile devices. We present a novel convolutional neural network (CNN) architecture to explicitly address the long standing problems in portrait segmentation, i.e., semantic coherence and boundary localization. Specifically, we propose a cross-granularity categorical attention mechanism leveraging the deep supervisions to close the semantic gap of CNN feature hierarchy by imposing consistent category-oriented information across layers. Furthermore, a cross-granularity boundary enhancement module is proposed to boost the boundary awareness of deep layers by integrating the shape context cues from shallow layers of the network. We further propose a novel and efficient non-parametric affinity model to achieve efficient instance segmentation on mobile devices. We present a portrait image dataset with instance level annotations dedicated to evaluating portrait instance segmentation algorithms. We evaluate our approach on challenging datasets which obtains state-of-the-art results.",
keywords = "Convolutional neural networks, Instance segmentation, Portrait segmentation, Semantic segmentation",
author = "Lingyu Zhu and Tinghuai Wang and Emre Aksu and Joni-Kristian K{\"a}m{\"a}r{\"a}inen",
year = "2019",
month = "7",
day = "1",
doi = "10.1109/ICME.2019.00281",
language = "English",
pages = "1630--1635",
booktitle = "2019 IEEE International Conference on Multimedia and Expo, ICME 2019",
publisher = "IEEE",

}

RIS (suitable for import to EndNote) - Lataa

TY - GEN

T1 - Portrait instance segmentation for mobile devices

AU - Zhu, Lingyu

AU - Wang, Tinghuai

AU - Aksu, Emre

AU - Kämäräinen, Joni-Kristian

PY - 2019/7/1

Y1 - 2019/7/1

N2 - Accurate and efficient portrait instance segmentation has become a crucial enabler for many multimedia applications on mobile devices. We present a novel convolutional neural network (CNN) architecture to explicitly address the long standing problems in portrait segmentation, i.e., semantic coherence and boundary localization. Specifically, we propose a cross-granularity categorical attention mechanism leveraging the deep supervisions to close the semantic gap of CNN feature hierarchy by imposing consistent category-oriented information across layers. Furthermore, a cross-granularity boundary enhancement module is proposed to boost the boundary awareness of deep layers by integrating the shape context cues from shallow layers of the network. We further propose a novel and efficient non-parametric affinity model to achieve efficient instance segmentation on mobile devices. We present a portrait image dataset with instance level annotations dedicated to evaluating portrait instance segmentation algorithms. We evaluate our approach on challenging datasets which obtains state-of-the-art results.

AB - Accurate and efficient portrait instance segmentation has become a crucial enabler for many multimedia applications on mobile devices. We present a novel convolutional neural network (CNN) architecture to explicitly address the long standing problems in portrait segmentation, i.e., semantic coherence and boundary localization. Specifically, we propose a cross-granularity categorical attention mechanism leveraging the deep supervisions to close the semantic gap of CNN feature hierarchy by imposing consistent category-oriented information across layers. Furthermore, a cross-granularity boundary enhancement module is proposed to boost the boundary awareness of deep layers by integrating the shape context cues from shallow layers of the network. We further propose a novel and efficient non-parametric affinity model to achieve efficient instance segmentation on mobile devices. We present a portrait image dataset with instance level annotations dedicated to evaluating portrait instance segmentation algorithms. We evaluate our approach on challenging datasets which obtains state-of-the-art results.

KW - Convolutional neural networks

KW - Instance segmentation

KW - Portrait segmentation

KW - Semantic segmentation

U2 - 10.1109/ICME.2019.00281

DO - 10.1109/ICME.2019.00281

M3 - Conference contribution

SP - 1630

EP - 1635

BT - 2019 IEEE International Conference on Multimedia and Expo, ICME 2019

PB - IEEE

ER -