AIVC: Artificial Intelligence based Video Codec

TL;DR

AIVC is an end-to-end neural video codec utilizing conditional autoencoders for efficient video compression, achieving competitive performance with HEVC and adaptable to various coding configurations.

Contribution

Introduces a novel neural video codec with autoencoder-based modules, enabling flexible compression and competitive performance with traditional standards.

Findings

Achieves performance comparable to HEVC in test conditions.

Demonstrates effective end-to-end rate-distortion optimization.

Provides comprehensive ablation studies on module contributions.

Abstract

This paper introduces AIVC, an end-to-end neural video codec. It is based on two conditional autoencoders MNet and CNet, for motion compensation and coding. AIVC learns to compress videos using any coding configurations through a single end-to-end rate-distortion optimization. Furthermore, it offers performance competitive with the recent video coder HEVC under several established test conditions. A comprehensive ablation study is performed to evaluate the benefits of the different modules composing AIVC. The implementation is made available at https://orange-opensource.github.io/AIVC/.