Optimization Accuracy of CNN Model by Utilizing CLAHE Parameters in Image Classification Problems

This study aims to optimize popular Convolutional Neural Network (CNN) models, such as ResNet50, InceptionV4, VGG19, MobileNetV1, MobileNetV2, MobileNetV3 Small, and MobileNetV3 Large, by utilizing the CLAHE (Contrast Limited Adaptive Histogram Equalization) parameter. This research focuses on the problem of classifying retinal fundus images, which have challenges in analyzing the structure and features of the picture. This study compared the performance of CNN models, which were optimized using the CLAHE method as image pre-processing. The fundus retinal image data used is a publicly available dataset. The optimization process is done by changing the parameters in the CNN model and applying the CLAHE technique to improve image quality before the classification process. The results showed that using the CLAHE parameter significantly contributed to improving the performance of the CNN model in classifying retinal fundus images. Some CNN models optimized with CLAHE produce better accuracy than the unoptimized models. There are also variations in performance between the different models, with some models providing better results in classifying retinal fundus images. This study provides new insights regarding using CLAHE parameters in optimizing the CNN model for retinal fundus image classification problems. The results of this research can be the basis for further development in medical image processing and pattern recognition for more accurate and practical diagnostic applications.