SceneDetector explains the scenario of Core ML -- using trained GoogLeNetPlaces model to detect scenes of a selected image. Literately 5 steps are required:

1. Drag and drop the .mlmodel to the project
2. Load model from the automatic generated model class of the .mlmodel file
3. Define a Vision service with the model
4. Create a request handler with image provided
5. Perform the request service with the request handler

For more details, please reference here.