| Abstrakt: |
Vision impairment, often caused by preventable ocular diseases can be challenging to diagnose accurately and prone to human error. Automation using technology, particularly deep learning, offers a promising solution to aid in accurate and efficient disease detection. This study explores the use of different CNN models specifically VGG-16, VGG-19, ResNet-50, and ResNet-152v2, for detecting ocular diseases. Simple fine-tuning is applied to these models, and their performance is compared to identify the most effective model. The purpose is to show how different models contribute to establishing reliable illness detection systems. The results reveal that most of these models perform well with even minimal fine-tuning. Among the models, ResNet-152v2 achieved the highest training accuracy of 90.36% demonstrating its capacity to learn from the training data. In contrast, ResNet-50 offered a more balanced performance with marginally lower accuracy, making it a robust choice for general application. [ABSTRACT FROM AUTHOR] |
