Ehdotuksen tiedot

This document specifies video coding technologies for machine consumption. Conventional video coding technologies aim for the best video under certain bit-rate constraints for human consumption. However, with the emergence of artificial intelligence systems, along with the abundance of sensors, many artificial intelligence platforms have been implemented with massive data requirements including scenarios such as connected vehicles, video surveillance, and smart city. The sheer quantity of data being produced constantly leads previous methods with a human in the pipeline to be inefficient and unrealistic in terms of latency and scale. There are additional concerns in transmission and archive systems which require a more compact data representation and low latency solution. This led to the introduction of the Video Coding for Machines standard specified in this document. In some cases, machines will communicate amongst themselves to perform tasks without a human in the mix, while in others there will be a need for additional human consumption of the specific decompressed stream. This specific scenario is possible in surveillance use cases, where a human “supervisor” occasionally searches for a specific person, or scene in video. In other cases, the corresponding bitstream is used for both human and machine consumption. The syntax, semantics, and decoding processes that are required to decode a bitstream are described in this document.