ryanwakabayashi / Ovarian_Cancer

Kaggle competition to detect Ovarian Cancer
0 stars 0 forks source link

Ovarian_Cancer

Kaggle competition to detect Ovarian Cancer

The data was downloaded from https://www.kaggle.com/competitions/UBC-OCEAN/data (737GB) and added to ./data within this project. A resized dataset can be found here for experimentation (2.54GB) https://www.kaggle.com/datasets/ryanwaka/ovarian-cancer-resized or downloaded using the following command: kaggle datasets download ryanwaka/ovarian-cancer-resized

The resized dataset folder was renamed and follows this path data/preprocessed_images/ train.csv will need to be downloaded as well from the kaggle link and added to the data/ directory

Data information:

Notes to look into:

Ideas to pursue:

Approaches:

  1. Naive - Preprocess and resize input images for a CNN

Future improvements:

  1. Allow argument parser for model selection

Training can be started by using: python train.py --model CNN

A new model can be added and trained by adding a .py file and a class within the file.

View Tensorboard:

  1. tensorboard --logdir runs
  2. Open http://localhost:6006/ in browser