Annotation Editor#
Knowing how to move around the annotation editor, learning all its capabilities and leveraging shortcuts will facilitate your annotation work. In this section, we will guide you through the annotation editor elements and describe the key concepts that drive annotation workflow.
Hint
Intel® Geti™ support for stylus
Annotating images is one of the functions of the Intel® Geti™ platform. The platform provides a user-friendly natural interface by using a stylus to enable users to directly draw on the image.
Key Concepts of the Annotation Workflow#
Active Learning#
One of the key features in the Intel® Geti™ platform is active learning, where the platform incrementally learns and actively suggests which images should be annotated next. This mechanism ensures that the model will be exposed to a wider variety of images and speed up the training process.
To learn more about active learning, read our AI fundamentals section.
Annotations vs. Predictions#
When a training round is finished and a model is available, the Intel® Geti™ platform can then draw the predictions coming from the model.
To distinguish user annotations and the predictions, the prediction labels are visualized differently. For predictions, you will see the following logo next to the label name 
. In addition, there is also a score next to the label name for predictions, depicted in percentage. This denotes how confident the model is in predicting that result. This score is typically referred to as confidence score; higher score means that the model is more confident with its predictions.
Note
Do not be alarmed with low confidence scores. As long as the location and the labels of the predictions shown in the screen fit the expectation, the model is performing well.
Annotation |
Prediction |
|---|---|
|
|
The user annotates four employees in the image. |
The model predicts that there are four employees in the image. The model has the following confidences in each prediction: 15%, 15%, 16%, and 10%. |
On the top-right corner of the annotation screen, there are two toggle buttons which can be used to quickly switch the visualisation between user annotation and model predictions.
User Annotation mode
- used for viewing annotations made by a userAI Prediction mode
- used for viewing annotations made by the system
User concurrency#
The Intel® Geti™ platform allows multiple users to annotate in a project concurrently. The system displays the name of the user who last annotated an item. If the user edits an existing annotation then the name of the latest editor appears on the annotation screen.