Eye is one of the prime sensory organs and illness in eye will cause mild to severe vision-related issues. The illness in eye is mainly due to infection or aging. Age-associated eye illness is one of the common issue and various medical protocols are available to detect and treat these illness. Cataract is one of the age-associated problem and early detection and treatment is essential. This research aims to propose a deep learning (DL) approach to classify the Retinal Fundus-image (RF) into normal/cataract. The various phases involved in the proposed scheme includes; (i) image collection and resizing, (ii) image enhancement using Otsu’s thresholding with Firefly Algorithm (OT + FA), (iii) feature extraction using NASNet model, and (iv) classification and threefold cross-validation. In this work, the performance of the proposed DL-tool is verified using the raw and processed RF and the experimental outcome of this study confirms that the NASNet-mobile model helps to provide a cataract detection accuracy of > 96% on the chosen data.