Back to Search
Document details
EE1-3.1: NNVC-LOP4 and LOP5 retraining with additional BVI-AOM dataset
Abstract
This contribution reports the results of EE1-3.1.1, EE1-3.1.2, EE1-3.1.3 and EE1-3.1.4 for LOP4 and LOP5 architectures. EE1-3.1.1 is the reproduction training of LOP4 adopted model with DIV2K, BVI-DVC and TVD datasets. EE1-3.1.2 is the training of LOP4 model with additional BVI-AOM dataset along with existing DIV2K, BVI-DVC and TVD datasets. Only stage 3 training was conducted for both the EE tests. EE1-3.1.3 is the reproduction training of LOP5 adopted model with DIV2K, BVI-DVC and TVD datasets. EE1-3.1.4 is the training of LOP5 model with additional BVI-AOM dataset along with existing DIV2K, BVI-DVC and TVD datasets. Only stage 3 training was conducted for both the EE tests.
JVET-AL0187 EE1-3.1: NNVC-LOP4 and LOP5 retraining with additional BVI-AOM dataset [T. Dumas, A. Monier, F. Galpin (InterDigital)] [late]
ARCHIVE
:
JVET-AL0187-v3.docx
NNVC-11-LOP4-int_vs_NNVC-12-LOP4-float.xlsm
NNVC-12-LOP4-float_vs_NNVC-12_LOP4-float_aom.xlsm
NNVC-12-LOP4-float_vs_NNVC-12_LOP4-float_idc.xlsm
nnvc_12_lop5_epochs_proportional_float_vs_nnvc_12_lop5_official_float.xlsm
nnvc_12_lop5_hyperparameters_unchanged_bs_decreased_float_vs_nnvc_12_lop5_official_float.xlsm
nnvc_12_lop5_hyperparameters_unchanged_bs_increased_float_vs_nnvc_12_lop5_official_float.xlsm
nnvc_12_lop5_hyperparameters_unchanged_float_vs_nnvc_12_lop5_official_float.xlsm
nnvc_12_lop5_hyperparameters_unchanged_lr_decreased_float_vs_nnvc_12_lop5_official_float.xlsm
nnvc_12_lop5_hyperparameters_unchanged_lr_increased_float_vs_nnvc_12_lop5_official_float.xlsm
nnvc_12_lop5_hyperparameters_unchanged_without_bvi_aom_float_vs_nnvc_12_lop5_official_float.xlsm
Decisions
JVET-AL0187 EE1-3.1: NNVC-LOP4 and LOP5 retraining with additional BVI-AOM dataset [T. Dumas, A. Monier, F. Galpin (InterDigital)] [late]
Citation