SiYtE situational awareness platform, using computer vision, IoT, edge computing, AI.

SiYtE situational awareness platform, using computer vision, IoT, edge computing, AI.

SiYtE is a situational awareness platform for complex and hazardous environments. It uses computer vision, IoT and third party data to automate the monitoring of these environments; provide workflow tools for alerting to resolve immediate issues; and capture longer-term trend analysis for performance monitoring, identification of issues and so on.

Its first customer was Network Rail. On the rail network the solution is used trackside for trespass detection, theft prevention at Redeployable Equipment Buildings, landslip monitoring, PPE detection for track workers, and other use cases. It is used on stations and concourses to monitor crowding, slips trips & falls, and incident detection eg vandalism. It is operational at Manchester Piccadilly, Leeds, Kings Cross St Pancras, as well as with MTR on the Elizabeth Line and in the South West. It is now finding adjacent market uses in Education (air quality monitoring with DfE), Healthcare (patient absconding with Nottingham UH Trust, and mental health service user behavioural monitoring with Lancs & South Cumbria FT), Ports (worker safety), and other sectors.

Architecturally SiYtE is set up to provide high levels of confidence (>95%), thus avoiding the false-positive issue that affects so many computer vision solutions. It deploys small models either on smart cameras or on Edge Compute if the existing camera network is not smart. When objects are identified & conditions are met to a defined accuracy (say 50%) a snapshot is taken & sent to the Purple cloud. There, larger inference models are run on the snapshot to validate - or not - that the conditions were satisfied to a higher accuracy (>95%), and that a workflow should be triggered or a data point captured. SiYtE can work with any existing camera install base - analogue through to smart cameras with onboard models. SiYtE is GDPR compliant, using facial blurring and redaction to prevent identification of individuals.