Organisation

• A: includes Task A code TaskA.py, all classes and functions required are already in TaskA.py

  • TaskA_ResNet.ipynb: can be run in Jupyter Notebook
  • Trained_Model.ipynb : file that can run the pretrained model
  • checkpoint_resnet100.pth: pretrained model with 100 epochs
  • loss.png: train loss curve used in the preort

• B: includes Task A code TaskB.py, all classes and functions required are already in TaskB.py

  • TaskB_ResNet.ipynb: can be run in Jupyter Notebook
  • Trained_Model.ipynb : file that can run the pretrained model
  • checkpoint_resnet100.pth: pretrained model with 100 epochs
  • loss.png: train loss curve used in the preort

• Dataset: includes directories of PathMNIST and PneumoniaMNIST. Bith directories have a temp.txt in it. pneumoniamnist.npz and pathmnist.npz can be put into corresponding directory.

• main.py: used for runing TaskA.py and TaskB.py

Package required

torch==2.1.1+cu118
torchvision==0.16.1+cu118
medmnist
sklearn
scipy
numpy
matplotlib
tqdm

To install medmnist as a standard Python package, use pip

    pip install medmnist

Or install from sorce

    pip install --upgrade git+https://github.com/MedMNIST/MedMNIST.git

Note

• The default training epoch is 10. It can also be changed in TaskA.py and TaskB.py with the parameter name of num_epoches