Introduction

This code is used to generate the command for model training automatically.

See more details from the official tutorial.

Usage

1. Setup the parameters in main.py to adapt to your dataset.
2. Automatically analyze the dataset, generate the model(s), and create the command used for model training by running:

python main.py

3. Setup your CUDA device ID, and train the model by running (FOR EXAMPLE):

bash segresnet_small_0.sh

Made the following changes before using it

• The default train_params is set as {}, instead of NONE. Otherwise, the code will raise the error "undefined parameters" at here.
• Add sys.path.insert(0, os.path.join(algo_path, "algorithm_templates")) to line 328 at here.
• Set Resample=False here for segresnet to avoid the out-of-memory issue. (This is NOT a bug. After modification, the default resolution is not (1, 1, 1) anymore.)
• Add extra code here to convert all the list-based transform to REAL transform objects
• Adapt to Windows operation system here

LOGS

• Add a small segresnet to assets where the initial_channel is set to 16, and the batch size is set as 1.