Global Deconvolutional Networks for Semantic Segmentation

Nekrasov, Vladimir; Ju, Janghoon; Choi, Jaesik

doi:10.48550/arxiv.1602.03930

Cited by 1 publication

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The use of deep convolutional neural networks for semantic segmentation is increasingly becoming popular since the recent success in dense object recognition [2,22,29]. Various techniques have been proposed to further improve the performance of dense prediction by deep neural networks, including global context information [15,25], learning deconvolution layers [26], applying conditional random fields as a post-processing [3,44], or incorporating weakly annotated data in training set [27]. In this paper, we propose a side-path encoder to predict unique set of consistent labels in segmentation and feed FCN output to fullyconnected CRF for addressing combinatorial preference issue in clothing parsing.…”

Section: Semantic Segmentationmentioning

confidence: 99%

Looking at Outfit to Parse Clothing

Tangseng,

Wu,

Yamaguchi

2017

Preprint

View full text Add to dashboard Cite

This paper extends fully-convolutional neural networks (FCN) for the clothing parsing problem. Clothing parsing requires higher-level knowledge on clothing semantics and contextual cues to disambiguate fine-grained categories. We extend FCN architecture with a side-branch network which we refer outfit encoder to predict a consistent set of clothing labels to encourage combinatorial preference, and with conditional random field (CRF) to explicitly consider coherent label assignment to the given image. The empirical results using Fashionista and CFPD datasets show that our model achieves state-of-the-art performance in clothing parsing, without additional supervision during training. We also study the qualitative influence of annotation on the current clothing parsing benchmarks, with our Web-based tool for multi-scale pixel-wise annotation and manual refinement effort to the Fashionista dataset. Finally, we show that the image representation of the outfit encoder is useful for dress-up image retrieval application.

show abstract