Deep Learning for Roman Handwritten Character Recognition

• DOI: 10.11591/ijeecs.v12.i2.pp455-460
• Publication Date: 2019-01-26T15:04:23Z
• Authors: Muhaafidz Md Saufi; Mohd Afiq Zamanhuri; Norasiah Mohammad; Zaidah Ibrahim
• Keywords: Similarity (geometry); Convolution (computer science)

The advantage of deep learning is that the analysis and massive amounts unsupervised data make it a beneficial tool for Big Data analysis. Convolution Neural Network (CNN) method can be used to classify image, cluster them by similarity, perform image recognition in scene. This paper conducts comparative study between three models, which are simple-CNN, AlexNet GoogLeNet Roman handwritten character using Chars74K dataset. produced results indicate GooleNet achieves best accuracy but requires longer time achieve such result while produces less accurate at faster rate.