RSNA-MICCAI Brain Tumor Radiogenomic Classification

Hardware/OS used for the competiion

• CPU: AMD Ryzen™ 7 5800X, 8cores
• GPU: RTX 3090
• RAM: 32GB
• OS: Ubuntu 20.04

Kaggle resources

The Kaggle competition link

Competition data

Solution Kaggle discussion

Inference notebook

Training steps

To train the model using the competition data, please follow these steps:

1. Clone this repo

2. Install dependencies via pip install -r requirements.txt

3. Download the competition data from here and place the downloaded files in the input folder.

4. Set the configuration file working/config.py. It is recommended to use the default values.

   • TRAINING_BATCH_SIZE: The training batch size
   • TEST_BATCH_SIZE: The testing/validation batch size
   • IMAGE_SIZE: The image size used during the training/infernece
   • N_EPOCHS: The number of the training epochs
   • NUM_IMAGES_3D: Number of the images/scans used to build the 3D images.
   • do_valid: bool that indicates if we want to save the model weights based on the validation score
   • n_workers: Number of workers used during the training

5. In case you want to use the pretrained weights from the final kaggle solution please skip the training step (next step). You will find the pretrained weights in weights/

6. Run training script bash working/train_valid.sh

7. Run prediction script python3 -m working.predict

If you have any questions, please don't hestiate to reach out to me on this email: [email protected]