Back to Search Document details
23rd Meeting: by teleconference, July 2021 2021-07-06 22:21
EE1: neural network based in-loop filter using depthwise separable convolution and regular convolution
Authors: Liqiang Wang Sheng Lin Xiaozhong Xu Shan Liu Cheung Auyeung Xinwei Li
Abstract
This contribution presents the updated result of JVET-V0137[1], the neural-network-based in loop filter using depthwise separable convolution (DSC) in EE1. With VTM11+ V0056[2][3] as anchor and test platform, on average results from RA/AI configurations, report gains for luma BD-Rate are 1.12% and 1.61% respectively.
Meeting Notes
JVET-W0151 EE1: Neural network based in-loop filter using depthwise separable convolution and regular convolution [L. Wang, S. Lin, X. Xu, S. Liu, C. Auyeung, X. Li (Tencent)] [late]

EE1 related contributions: Neural network-based video coding (6)

Contributions in this area were discussed in session 5 at 0500–0710 UTC on Thursday 8 July 2021 (chaired by JRO).

References:
JVET-T0057 JVET-U0060 JVET-U0061 JVET-V0137 JVET-V2016
Cited:
JVET-W0011 JVET-W0023 JVET-W0182 JVET-W2023 JVET-X0053
PATENTS:
WO/2024/010710 0.34 20220191553 0.33 WO/2022/132277 0.32 20220405545 0.31 WO/2022/261968 0.30 20220210446 0.28 WO/2022/146503 0.27 20210004589 0.26
ARCHIVE v1 :
JVET-W0151-EE-vtm11-MTCF_V0137.xlsm JVET-W0151-v1.docx
Decisions
Contributions in this area were discussed in session 5 at 0500–0710 UTC on Thursday 8 July 2021 (chaired by JRO).
Citation