Contextual ensemble network for semantic segmentation. (February 2022)

• Record Type: Journal Article
• Authors: Zhou, Quan; Wu, Xiaofu; Zhang, Suofei; Kang, Bin; Ge, Zongyuan; Jan Latecki, Longin
• Is Part Of: Pattern recognition. Volume 122(2022)
• Publication Date: 2022-02
• Subjects: Ensemble deconvolution -- Semantic segmentation -- FCNs -- Context aggregation -- Encoder-decoder networks; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2021.108290
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: Contextual ensemble network (CENet) introduces a novel encoder-decoder architecture to capture multi-scale context via ensemble deconvolution. The stacked feature maps are complemented each other, allowing us to fully explore multiple scale contextual cues embedded in images. Instead of using switch variables to record pooling indices, the feature maps of encoder are concatenated to the decoder, avoiding the extra noise introduced through padding, and without extra memory space to store pooling indices at the same time. The CENet is trained end-to-end and easy to execute without any post-processing, which facilitates well for semantic segmentation. The experimental results show the superior performance of CENet on CityScapes, PASCAL VOC 2012, MS COCO, and ISBI 2012 datasets. Abstract: Recently, exploring features from different layers in fully convolutional networks (FCNs) has gained substantial attention to capture context information for semantic segmentation. This paper presents a novel encoder-decoder architecture, called contextual ensemble network (CENet), for semantic segmentation, where the contextual cues are aggregated via densely usampling the convolutional features of deep layer to the shallow deconvolutional layers. The proposed CENet is trained in terms of end-to-end segmentation to match the resolution of input image, and allows us to fully explore contextual features through ensemble of dense deconvolutions. We evaluate our CENet on two widely-used semantic segmentation datasets: PASCAL VOC 2012 and CityScapes. The experimental results demonstrate our CENet achieves superior performance with respect to recent state-of-the-art results. Furthermore, we also evaluate CENet on MS COCO dataset and ISBI 2012 dataset for the task of instance segmentation and biological segmentation, respectively. The experimental results show that CENet obtains promising results on these two datasets.