Cataract Disease Diagnosis Using SURF Features and Pre-Trained Variants of an EfficientNet Model: Comparative Analysis

Cataracts are a common eye condition that affects millions of people worldwide. The major symptom of cataracts is poor or cloudy vision. Accurate and timely detection and diagnosis of cataracts can prevent vision loss. Therefore, an inexpensive system that diagnoses cataracts at an early stage needs to be developed. The goal of this paper is to use SURF features and pre-trained versions of an effective model trained on a publicly available image dataset to sort cataract disease into different groups. The variants of the EfficientNet model were applied to SURF features detected from the image dataset. The comparative analysis was conducted on the basis of different evaluation parameters and found that variant B0 has outperformed the rest of the variants of EfficientNet with an accuracy of 96.3%. The accuracy of EfficientNet B2 is greater than that of EfficientNet B1, but it can be considered to have better performance.