In order to identify and learn about anomalous patterns from distributed cameras with AI recognizers for various events, a team sought to leverage AI pattern identification across various ranges of time or impacting events.
The events triggered by the local recognizers (edge AI) are fed back to the Contextual platform and stored for ongoing processing and analysis. Events are categorized based on type and summarized across different time slices using Contextual Flows. Individual data points are further enriched based on local activities or happenings that may impact frequency or pacing of identification. Simple standard deviations are calculated based on segments of historic data and then new batches of data are submitted to LLM for analyzing and flagging anomalies in the data set inclusive of anticipated impact of local activities.
