With the introduction of the Facial Recognition module, we provide advanced backend image processing to accelerate the guest image retrieval operation, while adding convenience.

This well-tested system has processed tens of millions of images with a very high level of accuracy.

Our hybrid, multi-modal approach also employs other more traditional tagging methods that are used alongside facial recognition to further reinforce and enhance the superior CORTEX tagging capabilities.

The optimal workflow is to take the image and then have the server process the images during importing. This is the quickest, most reliant and certainly increases accuracy.

Once images are on the server, the image is first analyzed for face detection, then further refined to extract facial parameters, and then finally, each person in the image is processed through the advanced facial recognition algorithms.

Each detected facial contour pattern is then either:

1. enrolled as a new “subject” with a new subject ID, which is then available to verified against in the future, or
2. automatically compared and corroborated against the evolving cruise-specific database of facial contour architecture.

Once images are processed, a guest simply logs onto the sales system via the kiosk or mobile app and all images are automatically presented to them.

A wide array of regular cruise images will be recognized. On estimate, 85%+ of all regular vacation images are able to be processed.

In addition, the system is able to distinguish and process difficult situations, such as:
• Large groups of people
• Crowded environments
• Varied lighting conditions
• Partially obstructed features

The more images that are taken and tagged, the more accurate the results will become, however even a single processed image will return accurate results.

Lab Technicians can also manually process “orphaned” (un-recognized) images for second-level verification against an automatically generated declared set of images and subjects, which are returned when satisfying a confidence threshold for accuracy.

Whereas the system is extremely successful under normal parameters, we are working toward implementing 3D extrapolation and other cutting-edge processing methods to further enhance our system.

To address the present-day concerns of hacking and privacy, the processes and facial patterns are 100% secured on the encrypted database.

Guest representations, patterns and resulting data are not accessible to anyone, with a full data purge carried out after cruise end.

Each image that is processed is queried with an anonymous subject ID. All resulting correlated data between the image, ID and subject person then resides explicitly within the confines and purview of the cruise line’s guest manifest system architecture.