Exploring the Potential of InceptionResNetV2 for Robust Cataract Detection: A Deep Learning Perspective
Abstract
One important tool for visualizing things is an eye lens. It can occasionally be hazy, brought on by cataracts, and interfere with vision. To prevent blindness, cataracts must be identified and treated as soon as possible. Experts have put up several options in recent years for the early diagnosis of cataracts in the eyes. However, because early cataract identification is difficult, it takes a skilled ophthalmologist or eye surgeon to identify cataracts in their early stages. This is the age of artificial intelligence and digitization. AI is a topic that everyone is discussing. A lot of researchers suggest using AI to solve this issue. An AI-based cataract detection method was also suggested in this study. For this study, a binary class data set of eye fundus images was used. One well-known dataset for KAGGLE cataract detection is "Cataract." To improve the data set proficiency, a few picture augmentation techniques were applied. InceptionResNetV2 was used in this investigation, albeit with some hyper-parameter adjustments. Training accuracy of 99% and validation accuracy of 98% were attained using the suggested methods.