ABSTRACT
Conjunctivitis, commonly referred to as “pink eye,” is a prevalent and contagious eye condition that affects millions worldwide. Detecting conjunctivitis early and accurately is vital for timely intervention and effective management. Here, a convolutional neural network (CNN) model has been customized to automate the detection of conjunctivitis using eye images. Our dataset encompasses a diverse array of eye images, which include both healthy and conjunctivitis-affected cases. To tackle the challenge of limited data, we employ data augmentation techniques to expand the dataset. After pre-processing and augmentation, we curate a collection of 5135 eye images representing both pink-eye pathology and healthy states. Subsequently, these augmented images undergo classification using the developed CNN model. During execution, the customized CNN model obtains an impressive accuracy of 88.80%, with a loss of 0.25, and demonstrates precision, recall, and F1 scores of 0.50. The CNN model holds promise as an automated solution for conjunctivitis detection. Its accuracy and efficiency could substantially support medical professionals in early diagnoses, facilitating timely treatment and curbing transmission rates.
