Background Ophthalmic image quality is impacted by ocular pathology, decreasing accuracy of artificial intelligence (AI) algorithms classifying these images. This study uses AI to improve image quality on color fundus photos (CFPs) to determine if this will restore an AI algorithm’s accuracy in identifying diabetic retinopathy (DR) despite simulated cataracts (SCs). Methods A convolutional neural network (CNN) was trained to classify CFPs as either DR or non-DR. Cataracts were simulated by applying Gaussian blur, and the CNN was used to classify these images. A Very Deep Super-Resolution (VDSR) AI was used to enhance the images with simulated cataracts. The CNN then classified these images. Results The CNN was able to classify the dataset without any SC with an accuracy of 93%. This accuracy decreased to 78% with SC, but improved back to 85% after images were enhanced using the VDSR. Conclusion This study shows that despite AI accuracy decreasing with noise from SC, it can be improved partially using AI-based image enhancement.