Abstract: Preferred embodiments described herein relate to a pipeline framework that allows for customized analytic processes to be performed on multiple streams of videos. An analytic takes data as input and performs a set of operations and transforms it into information. The methods and systems disclosed herein include a framework (1) that allows users to annotate and create variable datasets, (2) to train computer vision algorithms to create custom models to accomplish specific tasks, (3) to pipeline video data through various computer vision modules for preprocessing, pattern recognition, and statistical analytics to create custom analytics, and (4) to perform analysis using a scalable architecture that allows for running analytic pipelines on multiple streams of videos.

Abstract: Preferred embodiments described herein relate to a pipeline framework that allows for customized analytic processes to be performed on multiple streams of videos. An analytic takes data as input and performs a set of operations and transforms it into information. The methods and systems disclosed herein include a framework (1) that allows users to annotate and create variable datasets, (2) to train computer vision algorithms to create custom models to accomplish specific tasks, (3) to pipeline video data through various computer vision modules for preprocessing, pattern recognition, and statistical analytics to create custom analytics, and (4) to perform analysis using a scalable architecture that allows for running analytic pipelines on multiple streams of videos.