We encourage the tracking of each phase 2 submission with a readme file.

Some of the things we encourage tracking include the following:

Model description:

-Model architecture, all intermediate layers and connections, details of activation, optimization, number of trainable parameters, loss function etc

- If not deep learning, provide model type and all relevant parameters

- Inputs/outputs : patches, 2D slice, 3D, modalities, etc

- Pre-processing of inputs

- Initialization of model parameters. If pre-trained on other existing datasets, specify which datasets

- Clearly state original work vs existing work.

- Existing code/software libraries/packages were used

Training method:

- Description of use of additional datasets (needs to be publicly available): data parameters and usage

- How was hyperparameter tuning performed, learning rate and batch size, other hyper parameters used

- Detail the data augmentation strategies

- Number of models trained if ensembles were used

- Model training time and estimated C02 impact