Overview:

• CNN: contains code of MFRNet, training and evaluation.
• CNN/models/VTM162_BD_MFRNet_ORI: contains MF-MFRNet models trained for 4 different QP settings.
• CNN/BC4/submit_train.sh: BC4 job script for training MFRNet.
• EVAL_BC4.sh : BC4 job script for decoding and evaluation.
• mainDec_EVAL.m: decoder MATLAB script, called in the job script above.
• Dockerfile: dockerfile used to generate the test environment on 3090 machines.

Requirements:

Supposed to run on BC4. Load the tensorflow module by running

module add languages/anaconda3/2019.07-3.6.5-tflow-1.14