Vivi Wulandari, Anggyi Trisnawan Putra

Optimization of the Convolutional Neural Network Method Using Fine-Tuning for Image Classification of Eye Disease

Introduction

Optimization of the convolutional neural network method using fine-tuning for image classification of eye disease. Optimize eye disease image classification using CNN and fine-tuning. This study shows fine-tuning with VGG16 improves accuracy from 82.63% to 94.13% for precise diagnosis.

Abstract

Abstract. The eye is the most important organ of the human body which functions as the sense of sight. Most people wish they had healthy eyes so they could see clearly about life around them. However, some people experience eye health problems. There are many types of eye diseases ranging from mild to severe. With advances in technology, artificial intelligence can be used to classify eye diseases accurately, one of which is deep learning. Therefore, this study uses the Convolutional Neural Network (CNN) algorithm to classify eye diseases using the VGG16 architecture as a base model and will be combined using a fine-tuning model as an optimization to improve accuracy. Purpose:To find out the accuracy results obtained in the fine-tuning optimization model on Convolutional Neural Network (CNN) method in classifying images in eye disease. Methods/Study design/approach: Combining the Convolutional Neural Network (CNN) method with fine-tuning optimization models for image classification in eye disease. The two methods will be compared to determine the best result. Result/Findings: The accuracy results obtained from testing the Convolutional Neural Network method with the VGG16 architecture were 82.63% while the accuracy results from testing the fine-tuning model were 94.13%. Novelty/Originality/Value: The test results on the fine-tuning model have better accuracy than the testing of the Convolutional Neural Network method. This can be seen in the fine-tuning model which has an increase in accuracy of 11.5%.

Review

The paper presents a focused study on optimizing Convolutional Neural Network (CNN) performance for image classification of eye diseases, a highly relevant application in medical diagnostics. The authors utilize the well-established VGG16 architecture as a base model and introduce fine-tuning as an optimization strategy. The primary objective is clearly stated: to evaluate the accuracy gains achieved through this fine-tuning process. The abstract outlines a straightforward comparison between a baseline CNN and its fine-tuned counterpart, indicating a clear methodological approach to address the research question. A significant strength of this work is the practical relevance of its application, aiming to leverage artificial intelligence for accurate eye disease classification. The choice of VGG16, a proven deep learning architecture, as the base model provides a solid foundation, and the subsequent implementation of fine-tuning is a well-regarded technique for enhancing transfer learning performance. The reported results are particularly compelling, showcasing a substantial increase in accuracy from 82.63% for the baseline CNN to an impressive 94.13% with fine-tuning, representing an 11.5% improvement. This finding strongly supports the effectiveness of the fine-tuning approach in this specific domain. While the abstract provides promising preliminary results, a thorough evaluation of the full paper would necessitate greater detail on several critical aspects. Crucially, information regarding the dataset used—its size, composition, specific eye disease categories, and image characteristics—is essential for assessing the generalizability and potential biases of the findings. Furthermore, a detailed description of the fine-tuning methodology, including hyperparameter settings, training procedures, and specific validation strategies, would be invaluable. Expanding beyond overall accuracy to include other performance metrics (e.g., precision, recall, F1-score, sensitivity, specificity) and a comparative analysis against other state-of-the-art models or architectures would further solidify the claims of novelty and superior performance.

Full Text

Comments

You need to be logged in to post a comment.