Group project about identifying the Coivd-19 from X-ray image
In the summer of 2020, I worked on a Deep learning project with Dr. Pavlos Protopapas who is the Scientific Director and Lecturer in Harvard University.
In that project I led a project aiming at figuring out people infected by COVID-19 by scanning and identifying lungs’ X-rays pictures.
- Our group constructed the “Xception Network” structure and used transfer learning to train our model. After searching for original documents, we also carefully tuned the parameters to improve our model’s performance to 85 percent.
- The whole project is running under the TensorFlow framework.
- To have a better understanding of the model, we extracted certain features of images with the increase of the depth of the network and utilized saliency maps to understand what the network was paying attention to.
This is the link to our presentation: Xception Network Covid Identify