OpenFARM provides a framework to allow easy integration of Media Analysis Components into media asset management (MAM) systems.
It relies in a single entry point (Manager) that manages Interpreters of Analysis Components, that can run locally or can be distributed across many boxes. This type of architecture, substantially increases the scalability of metadata Analysis Components tools (e.g. scene cut detection, text extraction) and allows a quick and easy integration with MAM systems.
Features:
* Written in Java, distributed across multiple machines using a client server framework. (Although everything can run on just the one machine).
* Server component can be deployed on a regular Java WebServer such as Tomcat or JBoss.
* The Framework is 'ontology-neutral'. Indexing metadata can be returned as strings and numbers, either singly or arrays. Other than that, there is no prescription for how each item/event is described.
* No new database required. All indexing meta-data is returned to the client application (the broadcast software), or cached until such time that the client asks for it. We figured that each Broadcaster has a specific configuration for data storage, and often a request for an extra database won't be well received.
* OpenFARM is designed to use multiple machines to run the analysis components. Indexing jobs can be easily added to the schedule, either through a web interface or through a web service. These jobs are load-balanced on the available machines.
* The analysis components can be written in (just about) any language (C, Java, python etc) and can be run on any operating system (linux, windows etc).
Comments