As Appeared on CNX Software:

MPEG Video Coding for Machines (VCM) is in the Works

A video codec for machines seems like a good topic for the first of April, or an article on the Onion. But based on a recent press release by Gyrfalcon Technology, this may become a real thing as the company partnered with China Telecom, and proposed a new video codec called “Video Coding for machines” (VCM) that provides compression coding for machine vision and human-machine hybrid vision.

Apparently a recent study published by Cisco in 2018, humans will become bit players in the “video watching business”, and Machine-to-Machine (M2M) applications will represent the greatest usage of Internet video traffic over the next four years. So the goal of the VCM group will be to establish a new standard that will improve the previous generation video coding and decoding standards such as H.264 (AVC), H.265 (HEVC) and H.266 (VVC).

Few details are provided so far, and I can’t find any VCM group in a web search. Obviously, this will not be for robots watching TV;), but more likely for AI and IoT applications that may use computer vision. Humans like to have plenty of  colors (10-bit/12-bit color depth) and resolution (1080p and 4K), but recent machine learning algorithms are often happy with 4-bit depth and fairly low resolutions (320×240 or even lower), so I can only assume VCM will be optimized for those use cases.

Beside China Telecom and Gyrfalcon, the VCM group will include Johanneum Research of Graz, Austria, Leibniz University of  Hannover, Germany, Peking University, Zhejiang University, the Institute of Computing Technology, as part of the Chinese Academy of Sciences, Huawei, ZTE, Lulu, Sony, NEC, Softbank, Honda, Samsung and LG.

If you want to follow the progress of the standard a mailing list has been setup for this purpose, and presentation entitled “Requirements of video analysis and semantic compression_Yuan ZHANG.pptx” was shared in one of the threads. I’ll see if I can get hold of it, and update the post when/if I do.