Weakly-Supervised Image Semantic Segmentation Based on Superpixel Region Merging
In this paper, we propose a semantic segmentation method based on superpixel region merging and convolutional neural network (CNN), referred to as regional merging neural network (RMNN). Image annotation has always been an important role in weakly-supervised semantic segmentation. Most methods use m...
Main Authors: | , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2019-06-01
|
Series: | Big Data and Cognitive Computing |
Subjects: | |
Online Access: | https://www.mdpi.com/2504-2289/3/2/31 |
id |
doaj-ebc9a58fa9d64452bfc809b89b2dc16a |
---|---|
record_format |
Article |
spelling |
doaj-ebc9a58fa9d64452bfc809b89b2dc16a2020-11-25T01:49:49ZengMDPI AGBig Data and Cognitive Computing2504-22892019-06-01323110.3390/bdcc3020031bdcc3020031Weakly-Supervised Image Semantic Segmentation Based on Superpixel Region MergingQuanchun Jiang0Olamide Timothy Tawose1Songwen Pei2Xiaodong Chen3Linhua Jiang4Jiayao Wang5Dongfang Zhao6Shanghai Key Lab of Modern Optical Systems, School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, ChinaDepartment of Computer Science and Engineering, University of Nevada, Reno, NV 89557, USAShanghai Key Lab of Modern Optical Systems, School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, ChinaInformation Science and Technology Research, Shanghai Advanced Research Institute, Chinese Academy of Sciences, No. 99 Haike Rd., Zhangjiang, Pudong, Shanghai 201210, ChinaShanghai Key Lab of Modern Optical Systems, School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, ChinaShanghai Key Lab of Modern Optical Systems, School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, ChinaDepartment of Computer Science and Engineering, University of Nevada, Reno, NV 89557, USAIn this paper, we propose a semantic segmentation method based on superpixel region merging and convolutional neural network (CNN), referred to as regional merging neural network (RMNN). Image annotation has always been an important role in weakly-supervised semantic segmentation. Most methods use manual labeling. In this paper, super-pixels with similar features are combined using the relationship between each pixel after super-pixel segmentation to form a plurality of super-pixel blocks. Rough predictions are generated by the fully convolutional networks (FCN) so that certain super-pixel blocks will be labeled. We perceive and find other positive areas in an iterative way through the marked areas. This reduces the feature extraction vector and reduces the data dimension due to super-pixels. The algorithm not only uses superpixel merging to narrow down the target’s range but also compensates for the lack of weakly-supervised semantic segmentation at the pixel level. In the training of the network, we use the method of region merging to improve the accuracy of contour recognition. Our extensive experiments demonstrated the effectiveness of the proposed method with the PASCAL VOC 2012 dataset. In particular, evaluation results show that the mean intersection over union (mIoU) score of our method reaches as high as 44.6%. Because the cavity convolution is in the pooled downsampling operation, it does not degrade the network’s receptive field, thereby ensuring the accuracy of image semantic segmentation. The findings of this work thus open the door to leveraging the dilated convolution to improve the recognition accuracy of small objects.https://www.mdpi.com/2504-2289/3/2/31superpixelCNNregion mergingSLICweakly-supervised |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Quanchun Jiang Olamide Timothy Tawose Songwen Pei Xiaodong Chen Linhua Jiang Jiayao Wang Dongfang Zhao |
spellingShingle |
Quanchun Jiang Olamide Timothy Tawose Songwen Pei Xiaodong Chen Linhua Jiang Jiayao Wang Dongfang Zhao Weakly-Supervised Image Semantic Segmentation Based on Superpixel Region Merging Big Data and Cognitive Computing superpixel CNN region merging SLIC weakly-supervised |
author_facet |
Quanchun Jiang Olamide Timothy Tawose Songwen Pei Xiaodong Chen Linhua Jiang Jiayao Wang Dongfang Zhao |
author_sort |
Quanchun Jiang |
title |
Weakly-Supervised Image Semantic Segmentation Based on Superpixel Region Merging |
title_short |
Weakly-Supervised Image Semantic Segmentation Based on Superpixel Region Merging |
title_full |
Weakly-Supervised Image Semantic Segmentation Based on Superpixel Region Merging |
title_fullStr |
Weakly-Supervised Image Semantic Segmentation Based on Superpixel Region Merging |
title_full_unstemmed |
Weakly-Supervised Image Semantic Segmentation Based on Superpixel Region Merging |
title_sort |
weakly-supervised image semantic segmentation based on superpixel region merging |
publisher |
MDPI AG |
series |
Big Data and Cognitive Computing |
issn |
2504-2289 |
publishDate |
2019-06-01 |
description |
In this paper, we propose a semantic segmentation method based on superpixel region merging and convolutional neural network (CNN), referred to as regional merging neural network (RMNN). Image annotation has always been an important role in weakly-supervised semantic segmentation. Most methods use manual labeling. In this paper, super-pixels with similar features are combined using the relationship between each pixel after super-pixel segmentation to form a plurality of super-pixel blocks. Rough predictions are generated by the fully convolutional networks (FCN) so that certain super-pixel blocks will be labeled. We perceive and find other positive areas in an iterative way through the marked areas. This reduces the feature extraction vector and reduces the data dimension due to super-pixels. The algorithm not only uses superpixel merging to narrow down the target’s range but also compensates for the lack of weakly-supervised semantic segmentation at the pixel level. In the training of the network, we use the method of region merging to improve the accuracy of contour recognition. Our extensive experiments demonstrated the effectiveness of the proposed method with the PASCAL VOC 2012 dataset. In particular, evaluation results show that the mean intersection over union (mIoU) score of our method reaches as high as 44.6%. Because the cavity convolution is in the pooled downsampling operation, it does not degrade the network’s receptive field, thereby ensuring the accuracy of image semantic segmentation. The findings of this work thus open the door to leveraging the dilated convolution to improve the recognition accuracy of small objects. |
topic |
superpixel CNN region merging SLIC weakly-supervised |
url |
https://www.mdpi.com/2504-2289/3/2/31 |
work_keys_str_mv |
AT quanchunjiang weaklysupervisedimagesemanticsegmentationbasedonsuperpixelregionmerging AT olamidetimothytawose weaklysupervisedimagesemanticsegmentationbasedonsuperpixelregionmerging AT songwenpei weaklysupervisedimagesemanticsegmentationbasedonsuperpixelregionmerging AT xiaodongchen weaklysupervisedimagesemanticsegmentationbasedonsuperpixelregionmerging AT linhuajiang weaklysupervisedimagesemanticsegmentationbasedonsuperpixelregionmerging AT jiayaowang weaklysupervisedimagesemanticsegmentationbasedonsuperpixelregionmerging AT dongfangzhao weaklysupervisedimagesemanticsegmentationbasedonsuperpixelregionmerging |
_version_ |
1725004652892651520 |