Back to Search Document details
39th Meeting: Daejeon, KR, March 2025 2025-03-19 16:26
EE1-3.1: Retraining LOP4 and LOP5 using extended dataset from BVI-AOM
Authors: Du Liu Ericsson J. Ström Ericsson Mitra Damghanian Ericsson P. Wennersten Ericsson
Abstract
This contribution reports the performance of retraining LOP4 and LOP5 using an extended training dataset from BVI-AOM. The training is performed for stage 3 only. The extended training dataset includes the current LOP training set and the additional sequences from BVI-AOM. Since the extended training dataset is larger, Test 3 reduces the number of epochs by 12% to match the number of steps used for LOP5.
Meeting Notes
JVET-AL0144 EE1-3.1: Retraining LOP4 and LOP5 using extended dataset from BVI-AOM [D. Liu, J. Ström, M. Damghanian, P. Wennersten (Ericsson)]
ARCHIVE :
JVET-AL0144_EE1-3.1.2_retrain_LOP4_float.xlsm JVET-AL0144_EE1-3.1.2_retrain_LOP4_int16.xlsm JVET-AL0144_EE1-3.1.4_retrain_LOP5_float.xlsm JVET-AL0144_EE1-3.1.4_retrain_LOP5_int16.xlsm JVET-AL0144_EE1-3.1.4_retrain_LOP5_reduced_epochs_float.xlsm JVET-AL0144_v1.docx
Decisions
JVET-AL0144 EE1-3.1: Retraining LOP4 and LOP5 using extended dataset from BVI-AOM [D. Liu, J. Ström, M. Damghanian, P. Wennersten (Ericsson)]
Citation