Skip to content

🧪 Testing Model Training

This page explains how to test the model training using the DL_Track_US GUI.

1. Before You Start

  • A working GPU is highly recommended; otherwise, model training will take significantly longer.

  • Instructions to setup the GUI and environment are found in the Installation Guidelines.

  • The test training images and masks you need are located in:

  • DL_Track_US_example/tests/model_training/

2. Important Setup Instructions

For this test, ensure the following:

  • Click on Advanced Methods and select "Train Model" in the dropdown menu.
  • Ignore the main GUI window for now — you will only use the Model Training window.
  • Use the correct training images:
  • DL_Track_US_example/tests/model_training/apo_img_example
  • Use the correct training masks:
  • DL_Track_US_example/tests/model_training/apo_mask_example
  • Keep the parameter settings exactly as shown.
  • Critical: Set the number of Epochs to 3 (for quick test training).

test_setup

3. Starting the Training

  • After setting all parameters, click Start Training.

  • During the process, you will encounter several messageboxes:

  • Confirm each by clicking OK.
  • These confirm that:
    • Images and masks have been loaded.
    • Model compilation was successful.
    • Training completed successfully.

4. After Training

Once training finishes, you should find three new files in your selected output folder:

  • 📄 Test_apo.xlsx — Training summary file
  • 📄 Test_apo.h5 — The trained model
  • 📄 Training_results.tif — A plot of the loss curve over epochs

⚠️ Note:
Because neural network training includes uncertainty, your results (e.g., final loss values) may slightly differ from ours.

If the three files are generated correctly, it means your DL_Track_US installation works properly for model training!