Training Your Model

Learn how to create versions of your ML model to begin training

Models in Sense are tracked by their specific model version. For all models other than custom models, adding a new version will begin a new round of training.

If you want to train your model with a data set, add a version and select the data set for training. You will be able to select which version you want to deploy to your cameras.

Only pre-trained models from the Sixgill Model Library can be re-trained with new data sets. Custom models are imported as-is and assigned to a model version. At this time, custom models cannot be enhanced with additional training data through Sense.

  • In the "ML" tab, select your model.

  • In your model's "Versions" tab, click "Add Version"

  • For pre-trained models

    • Select the training data set you wish to train this model with

    • Enter an optional "Version Description"

    • Click "Save Version & Train Model"

    • SmartML Training will begin. During the training run, it will progress through various Training States. The status will be Done when complete.

  • For custom models

    • Enter an optional "Version Description"

    • Enter the Dockerhub URL of the image that contains the model. If this is a private container, it must be accessible using the Dockerhub credentials entered in "Add Model"

    • If this is a GPU-enabled model, make sure "Model Requires GPU Resources" is checked

    • Click "Save Version"