This project aims to implement the code described in Article Number 1 and extends its capabilities using the findings from Article Number 2. The authors of Article Number 1 applied Transfer Learning on four different architectures to detect COVID-19 and classify chest X-ray images. In this project, we build upon the results presented in Article Number 2 by incorporating additional data from this dataset into the dataset of Article Number 1. Finally, we apply Transfer Learning on two specific architectures, Squeeznet and ResNet-18.
The project can be divided into the following key components:
We enhance the dataset used in Article Number 1 by incorporating additional data from this dataset. This expanded dataset will be used for training and evaluation.
We focus on one specific model architectures:
For the ResNet-18 model, we train only the last layer (Fully Connected) and the last sequential layer. This transfer learning approach allows us to leverage the pre-trained ResNet-18 model's features while adapting it to our specific task.