Working with custom models

You can save time and resources by using your own TensorFlow-based custom models with IBM® Maximo® Visual Inspection.

Custom models are also known as custom networks. The following software is installed for custom model support:

• TensorFlow 2.1.0
• Keras 2.3.1
• PyTorch 1.3.1 with Python 3.6 support

Custom models work similarly as other models in IBM Maximo Visual Inspection. However, be aware of the following differences:

• When you upload a custom model to the Custom Models page, you can use the model to train a data set in IBM Maximo Visual Inspection and generate an IBM Maximo Visual Inspection trained model.
• Custom models cannot be used to import pre-trained models into IBM Maximo Visual Inspection.
• Transfer learning is not supported for custom models.

For information about preparing an IBM Maximo Visual Inspection trained model by using a custom TensorFlow model, see Preparing custom models to train data sets.

For examples, detailed instructions, and sample files for using custom models, see https://github.com/IBMSystemsCenterMontpellier/visual-insights-custom-models.