Training your model

After you have created your datasets and labeled the images, you can train your model. Defect Detection App creates either a classification or segmentation model, depending on how you annotated the training images. Optionally, you can choose to create a heatmap model using the classifications you make in your dataset. For more information, see Heatmap model. As part of the training process, Defect Detection App uses a test dataset to evaluate the model. To create the test dataset, the dataset is split into a test dataset and a training dataset.

After training is complete, you can evaluate the performance of the model and make any necessary improvements. For more information, see Evaluating your model.

To view the existing model versions in the project, choose Model versions tab on the top navigation pane.

The following procedure shows you how to train your model.

To train your model

1. If you're not on the project details page, do the following:

   1. Sign in to the Defect Detection App Console.

   2. In the top navigation pane, choose Projects.

   3. On the projects page, choose the project. The console opens the project details page.

2. On the project details page, choose the Model versions tab

3. Choose Train new model version. The console raises a warning dialog box if your training or test dataset does not met the minimum requirements for training a model. Choose the Training dataset tab to review the requirements and add more images to your dataset.

4. If your datasets meet the requirements, the console opens the Train new model version dialog box.

5. Enter a Model name for the model.

6. If you want to create a model that generates a heatmap, choose Generate defect heat maps. For more information, see Heatmap model.

7. (Optional) Expand Advanced settings to select a higher image resolution to apply during training.

   Defect Detection App creates downscaled copies of your dataset images to use for training. Higher resolutions can result in better accuracy when detecting small defects, but result in longer training times and higher inference latency.

8. Choose Train to start training the model.

9. In the Model versions tab, you can see that training has started and you can see the current status in the Status column for the model. Training a model takes a while to complete.

10. Next step: When training is finished, see Evaluating your model.