Review of Deep Learning Using Convolutional Neural Network Model

Ari Kurniawan; Erlangga Erlangga; Tia Tanjung; Fenty Ariani; Yuthsi Aprilinda; Robby Yuli Endra

doi:10.4028/p-kzQ3XE

Paper Titles

Identification of Endophytic Fungi on the Leaves of Jeruju (Acanthus ilicifolius) which are Potential as Antibacteria
p.9

DNA Mutation of Rc Gene and the Phenotype of Aleurone and Pericarp of Local Red Rice from South Sumatera
p.15

Safe Avoidance Region Detection for Unmanned Aerial Vehicle Using Cues from Expansion of Feature Points
p.23

Development of the Obstacle Avoider of Fish Robot
p.31

The Implementation of Natural Language Processing in Manufacturing and Service Industry through Skateboard Monitoring Device
p.43

Review of Deep Learning Using Convolutional Neural Network Model
p.49

Risk Based Inspection Analysis on Primary and Secondary Cooling Pipe of the TRIGA 2000 Reactor
p.59

A Multifaceted Assessments of Occupational Hazards in a Building Construction Site: Insights from a Case Study
p.65

Settlement of Land Disputes between PT Selatan Agro Makmur Lestari (PT SAML) and Residents of Tirta Mulya Village, Air Sugihan District, Ogan Komering Ilir (OKI) Regency, South Sumatra Province, Indonesia
p.77

HomeEngineering HeadwayEngineering Headway Vol. 3Review of Deep Learning Using Convolutional Neural...

Review of Deep Learning Using Convolutional Neural Network Model

Abstract:

Machine Learning can be used to process a lot of data and learn patterns from that data to predict the future. One of the most widely used parts of machine learning is Deep Learning. The Deep Learning method that currently provides the most significant results in image recognition is Convolutional Neural Network (CNN). Convolutional Neural Network (CNN) is one of the deep learning algorithms used for computer vision use cases such as image or video classification and detecting objects within images or even image areas. Some research related to the CNN model states that this model has a very good accuracy of 92% but with a fairly small amount of data and the use of epochs, namely 100, resulting in a higher validation error value than the error value in the training process, so that over fitting will occur. Based on several problems in the related research literature, this article aims to identify the weaknesses and shortcomings of Deep Learning algorithms using CNN models that refer to the state of the art, so that they can be used as a reference for further research. The state of the art related to research in the last five years, the Deep Learning algorithm using the CNN model found that (1) The number of epochs can affect the accuracy of the CNN model, (2) 2. The application of architecture can affect the accuracy of the CNN model, (3) the application of the type of layer can affect the accuracy of the CNN model. Based on several problems in the research literature related to the identification of weaknesses and shortcomings of Deep Learning using the CNN model which refers to Table 1. State of the Art summary of literature review research for the last five years, it can be concluded that to increase the accuracy of the CNN model, it is necessary to increase the number of epochs, apply the right architecture according to the problems in the research conducted, and use the type of layer. The hypothesis of this article can be used as a reference for further research related to Deep Learning using the CNN model.