Histopathology-Cancer-Detection

https://github.com/HagenMarin/Histopathology-Cancer-Detection.git
Model files are in the Github repository and can also be found here: https://drive.google.com/drive/folders/1GIYPgyaLMV-HMPyDOagjZHL32Vt3XBfm?usp=sharing
used libraries:

Pytorch
Torchvision
pathlib
glob
cv2
pandas
pickle
scikit learn
numpy
matplotlib

Setup:

1. install all of the libraries listed above in your environment
2. clone the repository
3. download the data from https://www.kaggle.com/competitions/histopathologic-cancer-detection/data and extract everything into the folder of the repository
4. create 3 folders called "test_batches128", "train_batches128" and "valid_batches128"
5. run run_to_create_batches.py (those Batches will also be needed for the Collab)
6. if you get an error make sure all of the used libraries are installed in your env
7. create an empty folder called model

Training:
run train.py (you can use "python train.py -h" or "python train.py -help" to see all the commandline options)

Collab:
if you want to run the collab yourself you need to upload the folder "test_batches128" and the file "train_labels.csv" into your google drive (after running run_to_create_batches.py !!!)