Superpixel Image Classification with Graph Attention Networks

被引:41
作者
Avelar, Pedro H. C. [1 ,2 ]
Tavares, Anderson R. [1 ]
da Silveira, Thiago L. T. [3 ]
Jung, Cliudio R. [1 ]
Lamb, Luis C. [1 ]
机构
[1] Univ Fed Rio Grande do Sul, Inst Informat, Porto Alegre, RS, Brazil
[2] Data Sci Brigade, Porto Alegre, RS, Brazil
[3] Fed Univ Rio Grande, Ctr Computat Sci, Rio Grande, Brazil
来源
2020 33RD SIBGRAPI CONFERENCE ON GRAPHICS, PATTERNS AND IMAGES (SIBGRAPI 2020) | 2020年
关键词
D O I
10.1109/SIBGRAPI51738.2020.00035
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
This paper presents a methodology for image classification using Graph Neural Network (GNN) models. We transform the input images into region adjacency graphs (RAGs), in which regions are superpixels and edges connect neighboring superpixels. Our experiments suggest that Graph Attention Networks (GATs), which combine graph convolutions with self-attention mechanisms, outperforms other GNN models. Although raw image classifiers perform better than GATs due to information loss during the RAG generation, our methodology opens an interesting avenue of research on deep learning beyond rectangular-gridded images, such as 360-degree field of view panoramas. Traditional convolutional kernels of current state-of-the-art methods cannot handle panoramas, whereas the adapted superpixel algorithms and the resulting region adjacency graphs can naturally feed a GNN, without topology issues.
引用
收藏
页码:203 / 209
页数:7
相关论文
共 34 条
[1]   Superpixels and Polygons using Simple Non-Iterative Clustering [J].
Achanta, Radhakrishna ;
Susstrunk, Sabine .
30TH IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2017), 2017, :4895-4904
[2]   SLIC Superpixels Compared to State-of-the-Art Superpixel Methods [J].
Achanta, Radhakrishna ;
Shaji, Appu ;
Smith, Kevin ;
Lucchi, Aurelien ;
Fua, Pascal ;
Suesstrunk, Sabine .
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2012, 34 (11) :2274-2281
[3]  
[Anonymous], SLIC SUPERPIXELS SLI
[4]   FAUST: Dataset and evaluation for 3D mesh registration [J].
Bogo, Federica ;
Romero, Javier ;
Loper, Matthew ;
Black, Michael J. .
2014 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2014, :3794-3801
[5]  
Cohen T. S., 2018, ARXIV PREPRINT ARXIV
[6]  
Danel T., INT C NEUR INF PROC, P668
[7]   SplineCNN: Fast Geometric Deep Learning with Continuous B-Spline Kernels [J].
Fey, Matthias ;
Lenssen, Jan Eric ;
Weichert, Frank ;
Mueller, Heinrich .
2018 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2018, :869-877
[8]  
Hu J, 2018, PROC CVPR IEEE, P7132, DOI [10.1109/CVPR.2018.00745, 10.1109/TPAMI.2019.2913372]
[9]  
Huang YP, 2019, ADV NEUR IN, V32
[10]  
Jiang B., 2019, P IEEE C COMP VIS PA