Graph Convolutional Neural Networks for Computer Vision

12Case Study and Use Cases of Dynamic Graphs in GCNN for Computer Vision

S. Anubha Pearline* and S. Geetha

School of Computer Science and Engineering, Vellore Institute of Technology, Chennai Campus, Chennai, India

Abstract

In recent years, deep learning has been considered lucrative in various domains such as computer vision and natural language processing. Apart from deep learning, Graph Neural Networks (GNNs) have captured a lot of interest among researchers. Exclusively for computer vision applications, graph convolutional neural network (GCNN) is widely used. GNN and GCNN rely on graph databased neural networks. Convolutional neural networks (CNN) can have images as direct inputs, while GCNN reveals the image grids or patches as graph nodes. Many different GNNs and three various applications of computer vision are discussed in this book chapter.

Keywords: Graph convolutional neural networks, computer vision, dynamic graphs, convolutional neural networks

12.1 Introduction

Over the past decade, deep learning has risen and emerged as an outstanding technique in Artificial Intelligence (AI) and Machine learning (ML). The major highlight of deep learning is that it extracts a lot of innate features from images in Computer Vision (CV) applications [1]. Image processing or CV encompasses dataset acquisition, image pre-processing, extraction of features, and image classification [2]. Graph Convolutional Neural Networks (GCNNs) depend on graphs, nodes, and edges. According to ...