ayulockin / kagglerecipes

Helpful data preprocessing, training, and visualisation code and scripts for a range of Kaggle competitions, supported by Weights & Biases.

Kaggle Recipes

What's New

Sept 3, 2021

• Bug fix

Sept 2, 2021

• Release version 0.0.3

• Added support for multiprocessing to extract DICOM metadata for RSNA-MICCAI Brain Tumor Classification competition.

• Bug fixes.

Aug 27, 2021

• Released the library on PyPI.

• Easily create voxel manipulated dataset for RSNA-MICCAI Brain Tumor Classification competition.

• Extract dicom metadata.

• Added utilities to log dataframe as tables and files/directory as artifacts.

• Added utiities to log basic W&B charts (line, bar, and scatter).

Kaggle Competitions

• RSNA-MICCAI Brain Tumor Radiogenomic Classification

Code & Scripts

RSNA-MICCAI Brain Tumor Radiogenomic Classification

• LINK TO NOTEBOOK

Install

pip install kagglerecipes

Sample Datasets

We have also logged smaller subsets of Kaggle commpeition datasets local development and fast prototyping.

RSNA-MICCAI Brain Tumor Radiogenomic Classification

• Download it manually from here.

• Or download it using this code snippet.

  
  import wandb
  
  run = wandb.init()
  
  artifact = run.use_artifact('wandb_fc/rsna-miccai-brain/sample:v0', type='dataset')
  
  artifact_dir = artifact.download()
  
  

Credits

The code in this repository is a combination of many of the Kagglers' work that they have shared publicly as Kaggle kernels.

• Connecting voxel spaces by Michael Beregov

• Normalized Voxels: Align Planes and Crop by yu4u

• DICOM to 2D Resized Axial PNGs 256x256 [x36] by Sofia Moschou

Note that this is not an official Weights and Biases product.