GenAI Capabilities

Foundation models typically need higher end hardware to operate. That being said, the latest generation of AI accelerators provide significant capabilities at an affordable price point. 

Hybrid pipelines allows us to efficiently leverage AI resources by using traditional computer vision techniques in conjunction with foundation models. For example we may use traditional object detectors to first identify content of interest before sending it for analysis by a foundation model. 

Databases that support vector search are required to use GenAI capabilities. Our on-prem solution is based on OpenSearch and we have the option of using managed databases from all of the major cloud providers for hybrid installations. 

Our perspective is GenAI augments traditional video analytics capabilities instead of replace them. For example, well defined prescriptive events are often times better handled by CNN models while more abstract and free form tasks are better suited for GenAI foundation models.