Project

Overview

In my project, I conducted a comprehensive review of different supervised learning approaches to predict image results. After evaluating multiple models, I found that the transfer learning approach with VGG16 provided the best results. To enhance the dataset, I used data augmentation techniques and ensured that the model was not overfitting by using the same data split size for testing and validation. Ultimately, the transfer learning model achieved a 92% accuracy rate, as well as 91.7% precision and 94% recall.
In the final stages of the project, I incorporated the model into a web application to create a solution that can be used to predict pneumonia. This prototype application can be especially useful in emergency situations where there is a scarcity of radiologists, while still adhering to social, legal, and ethical compliance standards.