Fundus images are frequently utilized by medical professionals such as ophthalmologists and are quite useful in detecting various retinal abnormalities. They utilized this to diagnose a variety of eye diseases, including cataracts, diabetic retinopathy, and glaucoma. Automatic diagnosis of retinal diseases is a highly challenging task today. Retinal fundus images play a vital role in providing valuable information for ophthalmologists to the fast and accurate diagnoses. Early and accurate diagnosis of retinal diseases is crucial for timely intervention and treatment. The majority of people suffer from a lack of accurate diagnosis to prevent their vision loss. This research work employs a ResNet18 model to classify these fundus images into multi-labeled categories. Fundus images consist of four categories such as cataracts, diabetic retinopathy, glaucoma, and normal cases and these are major causes of vision impairment worldwide. The dataset comprises 4217 retinal fundus images belonging to four classes collected from Kaggle. Our proposed ResNet18 model is trained on this dataset, and we split the dataset into train-test-validation parts. The model has been trained & validated using Kaggle datasets. Finally, the proposed model, achieved 100 & 94% accuracy on the training and validation dataset.