AI Model Snapshots

AI model snapshots allow users to view and manage the progress of their models. Each snapshot represents the state of the model at a specific point in time, making it easy to track progress or run detection from an earlier state.

You can access the model's Training Snapshots Timeline from the ORCA HUNTR - MAP view by clicking on SNAPSHOTS as seen below.

Click here to open a list of the current models snapshots

The Training Snapshots Timeline provides you with a visual representation of the progress of your model. Each time the training process is stopped, a snapshot is automatically generated. Each snapshot generated in this list will be timestamped and include information on the number of training areas and scenes in use.

Here you can see, save and rename snapshots.

There are three types of snapshots in the timeline:

• Current state: This is the most recent snapshot of the model.
• Temporary Snapshot: Temporary snapshots are the default snapshots HUNTR creates for your model. These snapshots will be automatically deleted after they expire if they are not saved.
• Saved Snapshot: Saved snapshots will not be deleted automatically.

To convert a temporary snapshot to a saved one or to rename a saved snapshot click the ☰ icon for the desired snapshot.

All of the snapshots that appear in the Training Snapshots Timeline will be selectable when running detection.

Model Creation from Snapshot

You can create a new AI model from any existing model snapshot. To create a new model click the ☰ icon for the desired snapshot and select AI Model from Snapshot.

Create a model from an existing snapshot.

Model creation will have the same steps as other new models. However the created model must have the same number of classes as the existing model. The model will be created with all the training up to the snapshot used for creation.

You also have the option to reuse the training areas and labels from the selected snapshot. If you choose this option, the existing labels and training areas will continue to be used for new model training instead of being discarded.

Choose to reuse the existing labels and training areas.

Revert to Previous Training Snapshot

If you are dissatisfied with the progression of your model, you can revert the model to a snapshot that was a better fit. To revert a model to a previous snapshot, click the ☰ icon for the desired snapshot and select Revert to.

Revert a model to an existing snapshot.

The current state of the model will be reverted to this snapshot. Any training done after this snapshot will be discarded.

Warning

All progress and any training data generated after the training snapshot was created will be permanently lost!