CIS522-DL-Tribiotics
Team Members:
Trevor Chan, Jordan Matelsky, Jiazhen Rong
This repository contains our code for Upenn 2022 SPRING CIS 522 final project.
In this project, we classied public COVID-19 X-ray image dataset with ResNet and applied interpretable methods GRAD-CAM and SHAP to interpret the model and result.
Standard Training pipeline:
Including Data preprocessing & Transfer Learning of ResNet:
https://colab.research.google.com/drive/1gm7ZMR2nmQxS42vv4Osw4D6XJQWHbOZv?usp=sharing
(Anti masking training) - https://colab.research.google.com/drive/1SD2bzAuRItR6w8HCZgx4lprsx8KspsRr
Baseline visualization & GRAD-CAM pipeline:
Including Baseline CNN filters visualization & GRAD-CAM method:
https://colab.research.google.com/drive/1_7ubTDYWu7CzmX98T-S7cEpqbeaUjI7L?usp=sharing
SHAP pipeline & results:
https://colab.research.google.com/drive/1gA9Ti6JHqzEFF_uBe-ryWkiHOAlFc71a?usp=sharing
Machine Learning pipeline:
including logistic regression & random forest:
ML_Baseline.ipynb & Dataset_Exploration.ipynb within this github.
saved_models:
This folder contains trained ResNet weights for full X-ray image, keeping the lungs and masking out the lungs.