Back to Search Document details
42nd Meeting: Santa Eulària, ES, April 2026 2026-04-21 14:24
EE1-2.1: VLOP4 with new backbone block based on Spatial-Channel Mixing (SCM)
Authors: HyunDong Cho KHU S. Bahk KHU J. Kim KHU H. Y. Kim KHU Dong-Ha Kim ETRI Sung-Chang Lim ETRI
Abstract
This contribution reports EE1-2.1 results on the impact of a Spatial-Channel Mixing (SCM)-based backbone block for VLOP4. The proposed backbone block uses a depth-wise 3×3 convolution for spatial mixing and 1×1 pointwise convolutions that first expand and then reduce the channel dimension for channel mixing. The complexity and test results of EE1-2.1 are summarized as follows.
Meeting Notes
JVET-AP0201 EE1-2.1: VLOP4 with new backbone block based on Spatial-Channel Mixing (SCM) [H. Cho, S. Bahk, J. Kim, H. Y. Kim (KHU), D. Kim, S.-C. Lim (ETRI)]
ARCHIVE :
JVET-AP0201_v2.docx NNVC-16-VLOP vs Proposed-VLOP-int.xlsm NNVC-16-VLOP vs Proposed-VLOP-int_reparam.xlsm
Decisions
JVET-AP0201 EE1-2.1: VLOP4 with new backbone block based on Spatial-Channel Mixing (SCM) [H. Cho, S. Bahk, J. Kim, H. Y. Kim (KHU), D. Kim, S.-C. Lim (ETRI)]
Citation