MultiStep Regression  Classification for Employee Attrition  Salary Estimation

Diabetic Retinopathy (DR) is a progressive eye disease caused by diabetes, and early detection is critical to prevent vision loss. This project, titled "Efficient SVM: A Hybrid Model for Diabetic Retinopathy Classification Using Retinal Fundus Images", presents a hybrid deep learning and machine learning framework for the accurate classification of DR stages. Leveraging pre-trained MobileNet for efficient feature extraction and Support Vector Machine (SVM) for robust classification, the system aims to balance computational efficiency with high predictive performance. The model is trained on the APTOS 2019 Kaggle dataset comprising labeled retinal fundus images across five classes: No_DR, Mild_DR, Moderate_DR, Severe_DR, and Proliferative_DR. The image processing pipeline includes resizing, normalization, and feature vector extraction using MobileNet, followed by classification via SVM. Additionally, the system is deployed through a Flask web interface, allowing users to upload retinal images and receive immediate DR classification results. The modular architecture supports model expansion to include EfficientNet-SVM and transformer-based classifiers such as ViT and Swin Transformer for performance comparison and enhancement. This approach demonstrates the effectiveness of transfer learning and hybrid modeling in medical image classification, offering a scalable solution for real-time, low-latency DR diagnosis. Future work will integrate explainable AI modules for clinical interpretability and extend model capabilities for broader ophthalmic disorder detection. Keywords
Diabetic Retinopathy, Hybrid Model, MobileNet, SVM, Retinal Fundus Images, Deep Learning, Machine Learning, EfficientNet, Swin Transformer, ViT, Transfer Learning, Flask Deployment, Medical Image Classification.