CNN-G: Convolutional Neural Network Combined With Graph for Image Segmentation With Theoretical Analysis

Lu, Yi; Chen, Yaran; Zhao, Dan; Bao, Liu; Lai, Zhichao; Chen, Jianxin

doi:10.1109/tcds.2020.2998497

Cited by 49 publications

(18 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A dilated convolution strategy was presented in [44] for replacing the pooling strategy in the conventional convolutional neural network, which reduces the resolution reduction resulted from the pooling operation. In [27], authors theoretically analyzed the CNN-G framework combining convolutional neural network with graph for semantic image segmentation application. In the community of SAR image processing, a joint convolutional neural network was presented for simultaneous despeckling and classification of SAR targets in [21].…”

Section: Cnn Based Methodsmentioning

confidence: 99%

SAR image segmentation with parallel region merging

Zhang

Xiong

et al. 2020

Multimed Tools Appl

View full text Add to dashboard Cite

In this paper, a parallel region merging strategy is proposed for partitioning Synthetic Aperture Radar (SAR) images into several disjoint regions, based on the region adjacency graph (RAG) of an initial partition and the nearest neighbor graph (NNG) produced from the RAG. Developed from the multi-direction ratio edge detector, a multi-scale-multi-direction (MSMD) one is used to extract edge strength map (ESM) of an initial SAR image, feeded into watershed transform to generate an initial partition result of the initial SAR image. Considering local image properties, which makes the generated NNG center around binode circles situated in interiors of homogeneous regions, many of which are independently located in different homogeneous regions, the predication of the parallelizability for bi-node circles is proposed to make the proposed parallel region merging strategy. The proposed parallel merging strategy simultaneously merges bi-node circles far away from boundaries of regions, characterized by the length of path from a node to the bi-node circle in the NNG. The performance of the proposed parallel merging strategy is analyzed theoretically and experimentally, and our experiments show that the proposed method outweighs other compared methods.

show abstract

Section: Cnn Based Methodsmentioning

confidence: 99%

SAR image segmentation with parallel region merging

Zhang

Xiong

et al. 2020

Multimed Tools Appl

View full text Add to dashboard Cite

show abstract

“…Here, different kinds of segmentation loss are discussed including multi-class cross-entropy loss, generalised dice loss, and their combination. The definition of multi-class cross-entropy loss can be computed by Equation (3).…”

Section: B Loss Functionmentioning

confidence: 99%

“…The superiority of feature extraction helps DNN based approaches to make compelling achievement on semantic segmentation. Deep learning based image segmentation model [3,4] has been utilised to understand the surrounding traffic environment of the autonomous vehicle to enhance its driving safety. When deploying such a model, each pixel of an image is assigned to one of the semantic classes, such as car, truck, tree, or pedestrian.…”

Section: Introductionmentioning

confidence: 99%

Improving Synthetic to Realistic Semantic Segmentation With Parallel Generative Ensembles for Autonomous Urban Driving

Fang

Hua

et al. 2022

IEEE Trans. Cogn. Dev. Syst.

View full text Add to dashboard Cite

Semantic segmentation is paramount for autonomous vehicles to have a deeper understanding of the surrounding traffic environment and enhance safety. Deep neural networks (DNN) have achieved remarkable performances in semantic segmentation. However, training such a DNN requires a large amount of labelled data at pixel level. In practice, it is a labour-intensive task to manually annotate dense pixel-level labels. To tackle the problem associated with a small amount of labelled data, Deep Domain Adaptation (DDA) methods have recently been developed to examine the use of synthetic driving scenes so as to significantly reduce the manual annotation cost. Despite remarkable advances, these methods unfortunately suffer from the generalisability problem that fails to provide a holistic representation of the mapping from the source image domain to the target image domain. In this paper, we therefore develop a novel ensembled DDA to train models with different upsampling strategies, discrepancy and segmentation loss functions. The models are, therefore, complementary with each other to achieve better generalisation in the target image domain. Such a design does not only improve the adapted semantic segmentation performance, but also strengthen the model reliability and robustness. Extensive experimental results demonstrate the superiorities of our approach over several state-of-the-art methods.

show abstract

“…The basic idea of image manipulation detection is to capture the tampering artifacts to identify and localize the manipulated regions for image authenticity determination. The traditional manipulation detection approaches mainly extract handcrafted features to identify and localize manipulated regions, such as Color Filter Array (CFA) [Popescu and Farid, 2005], Discrete Cosine Transform (DCT) [Li et al, 2017], Scale-Invariant Feature Transform (SIFT) [Zhou et al, 2016]. Such approaches have achieved impressive performance.…”

Section: Introductionmentioning

confidence: 99%

Learning Hierarchical Graph Representation for Image Manipulation Detection

Wang¹,

Zhou²,

Ling³

et al. 2022

Preprint

View full text Add to dashboard Cite

The objective of image manipulation detection is to identify and locate the manipulated regions in the images. Recent approaches mostly adopt the sophisticated Convolutional Neural Networks (CNNs) to capture the tampering artifacts left in the images to locate the manipulated regions. However, these approaches ignore the feature correlations, i.e., feature inconsistencies, between manipulated regions and non-manipulated regions, leading to inferior detection performance. To address this issue, we propose a hierarchical Graph Convolutional Network (HGCN-Net), which consists of two parallel branches: the backbone network branch and the hierarchical graph representation learning (HGRL) branch for image manipulation detection. Specifically, the feature maps of a given image are extracted by the backbone network branch, and then the feature correlations within the feature maps are modeled as a set of fully-connected graphs for learning the hierarchical graph representation by the HGRL branch. The learned hierarchical graph representation can sufficiently capture the feature correlations across different scales, and thus it provides high discriminability for distinguishing manipulated and nonmanipulated regions. Extensive experiments on four public datasets demonstrate that the proposed HGCN-Net not only provides promising detection accuracy, but also achieves strong robustness under a variety of common image attacks in the task of image manipulation detection, compared to the state-of-the-arts.

show abstract

CNN-G: Convolutional Neural Network Combined With Graph for Image Segmentation With Theoretical Analysis

Cited by 49 publications

References 28 publications

SAR image segmentation with parallel region merging

SAR image segmentation with parallel region merging

Improving Synthetic to Realistic Semantic Segmentation With Parallel Generative Ensembles for Autonomous Urban Driving

Learning Hierarchical Graph Representation for Image Manipulation Detection

Contact Info

Product

Resources

About