I2-Multimedia-Lab / Dilated-CNN-based-deep-reference-picture-generation

Dilated convolutional neural network-based deep reference picture generation for video compression (ICASSP 2022)

Home Page:https://arxiv.org/abs/2202.05514

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

OurProject

This is the implemention of the paper "DILATED CONVOLUTIONAL NEURAL NETWORK-BASED DEEP REFERENCE PICTURE GENERATION FOR VIDEO COMPRESSION".

Paper URL: https://arxiv.org/abs/2202.05514

Requirement

python = 3.7.6
pytorch = 1.8.0
torchvision = 0.10.0
cudatoolkit = 10.2
cudnn = 7.6.5
VTM = 10.0

Note

File EncGOP.cpp is a modification of VTM inter-frame coding, the purpose is to add the deep reference picture to the reference picture list.

About

Dilated convolutional neural network-based deep reference picture generation for video compression (ICASSP 2022)

https://arxiv.org/abs/2202.05514


Languages

Language:C++ 71.9%Language:Python 28.1%