First, to explain the terms used: Edge in this sense could be computing hardware connected to an existing security camera with a view of the floor, or a self-checkout running a Windows or Linux operating system; in fact, anywhere that people interact with your business and there is capability to watch that engagement with computer vision.
Cloud here refers to the computer capability in a retailer’s own datacentre, or a commercial 3rd-party, such as Microsoft Azure, Amazon Web Services or Google Cloud.
‘Working at the edge’ means part or all of the SeeWare solution can run on the edge hardware, taking advantage of local resource. Data from the edge can be images and/or text data that describes elements of the camera video stream. This can be combined with data from other sensors, such as barcode readers and scales to output insights in simple text data to aid business logic that runs in the cloud.
From detecting spills to identifying suspicious activity at the checkout, SeeWare delivers a comprehensive solution from edge to cloud.
For our retail partners, it is important that data is generated and actioned in real-time; a spill must be identified, and a clean-up team dispatched as quickly as possible; any suspicious activity at a self-checkout should immediately generate an alert to either prompt the shopper to rescan or request help from a store assistant.
SeeChange have developed SeeWare to be a platform that enables the most effective methods to bring the benefits of AI/ML (Artificial Intelligence and Machine Learning) to your enterprise.
SeeWare is built as a set of functional components in containers (see our blog post SeeChange and containers). Use of containers in our edge-to-cloud design means that we can run each component where it is most effective and efficient, and automatically scale to handle more events from more camera streams.
The sequence of actions for any AI vision system will include:
- Processing of the camera stream – decode, crop, contrast, colour balance, sharpness
- Identification of objects in each frame – background, static objects, random non-retail items
- Redaction – removal or obscuration of personally identifiable elements to ensure compliance with GDPR and personal data privacy
- State engine – to follow a flow of expected actions and identify the exceptions
- Data analysis – reduction of inference data to essential data needed for business intelligence
- Interface – SeeWare interfaces with your business intelligence solution via a secure API (Application Programming Interface) – a standard set of commands and responses
The logic and function of the process is handled by SeeWare and runs on hardware that is appropriate and available for each task, with secure, encrypted communication at each stage. SeeWare logs processes in each container so that every stage can be tracked. This provides key information to enable incremental improvement. Our aim is to deliver the most cost-effective solution for our customer.
Here are a couple of examples of how the SeeWare AI vision solution can be delivered:
- Spill Detect can run completely on servers in the same datacentre that runs the VMS (Video Management System), handling thousands of camera streams from premises across a region.
- Our fresh produce solution runs on each SCO (self-checkout), with the ability to ‘learn’ new products on one SCO that can then be added to the product list on all SCO in the store and updated to all stores in region through a secure cloud platform.
Please get in touch with us at SeeChange Technologies to discuss your AI vision requirements.
