Back to Search Document details
16th Meeting: Geneva, October 2019 2019-10-01 17:07
JVET AHG report: Neural Networks in Video Coding (AHG9)
Authors: Shan Liu Y.M. Li Byung-Yoon Choi K. Kawamura Yun Li Liqiang Wang Ping Wu Hyeong-Gyun Yang
Abstract
This document summarizes the activity of AHG9: Neural network in video coding between the 15th meeting in Gothenburg, SE (3-12 July 2019) and the 16th Meeting at Geneva, CH (1-11 October 2019).
Meeting Notes
JVET-P0009 JVET AHG report: Neural Networks in Video Coding (AHG9) [S. Liu, Y.M. Li, B. Choi, K. Kawamura, Y. Li, L. Wang, P. Wu, H. Yang]

This AHG report was discussed Thursday 3 October (chaired by JRO).

There was no relevant email activity on the reflector during this period.

One AHG9 related input document for this meeting was identified:

  • H. Yin, R. Yang, X. Fang, Z. Gao, R. Yang, “AHG9: Multiple Convolution Neural Networks For Sequence-Independent Processing,” JVET-P0489

The AHG recommendsed:

  • To review all related contributions
  • To continue investigating the benefits and complexity of using neural networks in video coding

It was suggested that the AHG has fulfilled its mandates by studying the potential of NN technology over several meeting cycles. Currently, there is no evidence that non-shallow NN based technology provides a good tradeoff of complexity vs. compression (at least w.r.t. needs of normative elements).

At this point of VVC development, and the low amount of recent contributions, it was suggested to discontinue the AHG.

References:
JVET-P0489
ARCHIVE v2 :
JVET-P0009.docx
Decisions
At this point of VVC development, and the low amount of recent contributions, it was suggested to discontinue the AHG.
Citation