Train and Predict

Train a model using your labeled data and make predictions on new images.

🏋️ Training & Prediction Panel

Steps:

1️⃣ Choose a model from dropdown and configure parameters.
2️⃣ Press 'Train network' a popup will appear that allows you to configure additional parameters.
3️⃣ Hit 'Predict current image' or 'Predict all images' to use the trained model to predict labels.

Training popup

After hitting train a popup should appear which allows you to further adjust training parameters.

Saved model and training data

After training you should find a saved model in the models directory. The exact format will depend on the frameworks used (for example in the screenshot below a monai pytorch model has been saved as a pth file, other frameworks may save in different formats). Some frameworks also output training statistics. In the below screenshot training statistics are saved in a .csv file.

After training (or after loading or setting a model) choose Predict current image or Predict all images to predict. The prediction layer should be populated with the predictions as shown in the below screen shot.

After predicting you need to save the project again and the predictions will be written to disk

The predictions will be written in your project folder under predictions\class_0.

---

🔄 Next: Run & Export