Abstract: This project develops a deep learning system for classifying medical images using the MobileNetV2 architecture and transfer learning. The images are preprocessed with resizing and normalization, and the model is trained for 25 epochs with custom layers to improve accuracy and F1 score. It also includes a Gradio interface for real-time predictions, enhancing diagnostic efficiency. The tool is designed to support medical professionals in areas with limited resources by offering a reliable and accessible diagnostic solution.
Index Terms—Deep Learning, Medical Image Classification, MobileNetV2, Transfer Learning, Image Preprocessing, Gra-dio Interface, Real-Time Predictions, Resource-Limited Settings, Glaucoma Detection, F1-Score, Diagnostic tool.

Keywords: Detecting Pediatric Glaucoma

| DOI: 10.17148/IARJSET.2025.125182