Performance Evaluation of Deep Learning Models for Image Enhancement and Classification in Medical Imaging Applications

This research for photo augmentation and classification, four deep learning models VGG InceptionNetV2, InceptionV3, MobileNetV2, and VGG16—are compared. For models, Accuracy, Precision, and F1 Score were used to measure how well they worked. PSNR, Entropy, SSIM, and MSE were some of the standards for picture quality. InceptionV3 made images look much better by lowering the MSE to 905.35, raising the PSNR to 18.56 dB, raising the SSIM to 0.625, and raising the Entropy to 6.40 %. These tests show that the contrast and characteristics of the image were kept. The fact that VGG16 has a higher accuracy (0.925) and F1 Score (0.916) shows that it is better at finding and classifying important features. VGInceptNetV2 greatly reduced false positives, with a maximum precision of 0.922. MobileNetV2 did not do well at augmentation or categorization. This comparison demonstrates that InceptionV3 is better for applications that need high-quality images and VGG16 is better for jobs that need high-accuracy classification. This shows the tradeoff between image quality and classification accuracy. This project enhances AI-driven imaging systems and develops more intelligent visual monitoring and analytical tools to enhance the safety and sustainability of cities and communities, in accordance with Sustainable Development Goals 9 (Industry, Innovation, and Infrastructure) and 11 (Sustainable Cities and Communities).