Skip to main navigation Skip to search Skip to main content

Fast portrait matting using spatial detail-preserving network

  • Shaofan Cai
  • , Biao Leng*
  • , Guanglu Song
  • , Zheng Ge
  • *Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Image matting plays an important role in both computer vision and graphics applications. Natural image matting has recently made significant progress with the assistance of powerful Convolutional Neural Networks (CNN). However, it is often time-consuming for pixel-wise label inference. To get higher quality matting in an efficient way, we propose a well-designed SDPNet, which consists of two parallel branches—Semantic Segmentation Branch for half image resolution and Detail-Preserving Branch for full resolution, capturing both the semantic information and image details, respectively. Higher quality alpha matte can be generated while largely reducing the portion of computation. In addition, Spatial Attention Module and Boundary Refinement Module are proposed to extract distinguishable boundary features. Extensive Experiments show that SDPNet provides higher quality results on Portrait Matting benchmark, while obtaining 5x to 20x faster than previous methods.

Original languageEnglish
Title of host publicationNeural Information Processing - 25th International Conference, ICONIP 2018, Proceedings
EditorsLong Cheng, Andrew Chi Sing Leung, Seiichi Ozawa
PublisherSpringer Verlag
Pages329-339
Number of pages11
ISBN (Print)9783030042233
DOIs
StatePublished - 2018
Event25th International Conference on Neural Information Processing, ICONIP 2018 - Siem Reap, Cambodia
Duration: 13 Dec 201816 Dec 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11306 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference25th International Conference on Neural Information Processing, ICONIP 2018
Country/TerritoryCambodia
CitySiem Reap
Period13/12/1816/12/18

Keywords

  • Deep learning
  • Detail-preserving
  • Fast matting
  • Portrait

Fingerprint

Dive into the research topics of 'Fast portrait matting using spatial detail-preserving network'. Together they form a unique fingerprint.

Cite this