The correct option is Performing inference. This is the process used when a trained model at a live endpoint analyzes a new catalog photo and returns labels or detections.

Performing inference means applying a trained model to new inputs so it produces predictions such as classifications or bounding boxes. The startup calls its deployed model with the photo and the model performs inference to identify the products in the image.

Model deployment is about hosting and exposing the model so it can be called. Deployment happens before predictions are served and it does not describe the act of producing labels.

Training the model is the stage where the model learns parameters from labeled data through repeated updates. Training is not the step where the trained model is used to predict new samples.

Amazon Rekognition is a managed computer vision service provided by AWS and it is not the name of the process. The question asks for the name of the action performed by the deployed custom model which is inference.

Full AWS Practitioner Certification Question