Back to Search Document details
29th Meeting: by teleconference, DE, January 2023 2023-01-17 22:00
EE1-1.6: NN chroma model without partitioning input
Authors: J. Ström Ericsson Du Liu Ericsson K. Andersson Ericsson P. Wennersten Ericsson Mitra Damghanian Ericsson Ruoyang Yu Ericsson
Abstract
This contribution presents results for a modification of the Neural Network loopfilter filter set 1, where the partitioning input to the intra chroma network has been removed. It is reported that the intra chroma model has been retrained using the same training procedure as for NNVC-3.0 filter set 1. The reported BDR results compared to NNVC3.0 filter set 1 are (SADL, int16):
Meeting Notes
JVET-AC0143 EE1-1.6: NN chroma model without partitioning input [J. Ström, D. Liu, K. Andersson, P. Wennersten, M. Damghanian, R. Yu (Ericsson)]

No need for presentation – th adopted EE1-1.5 removes the intra model entirely there is no need to discuss removing certain inputs for the chroma intra model.

Cited:
JVET-AC0234 X
ARCHIVE :
._JVET-AC0143-v2.docx ._JVET-AC0143.pptx ._JVET-AC0143_simplified_chroma_over_NNVC-3.0_filter_set_1.xlsm ._JVET-AC0143_simplified_chroma_over_NNVC-3.0_noNNLF.xlsm ._~$ET-AC0143-v2.docx JVET-AC0143-v2.docx JVET-AC0143.pptx JVET-AC0143_simplified_chroma_over_NNVC-3.0_filter_set_1.xlsm JVET-AC0143_simplified_chroma_over_NNVC-3.0_noNNLF.xlsm ~$ET-AC0143-v2.docx
Decisions
adopted
No need for presentation – th adopted EE1-1.5 removes the intra model entirely there is no need to discuss removing certain inputs for the chroma intra model.
Citation