How to Design and Train Your Implicit Neural Representation for Video Compression

2025-06-30

AI Summary

Implicit Neural Representations (INRs) for video compression suffer from low coding efficiency due to frame-wise independent training, hindering practical deployment. To address this, we systematically decouple the NeRV architecture and propose RNeRV—a lightweight, efficient model incorporating hypernetwork-based weight prediction and learnable weight masking. These innovations enable single-pass training with adaptive bit-rate control. Evaluated on seven 1080p videos, RNeRV achieves a 1.27% average PSNR gain with only a 0.4% parameter increase. On UCF-101 at 0.037 bpp, it improves PSNR and MS-SSIM by 1.7% and 2.5%/2.7%, respectively. This work significantly advances INR-based video compression toward real-time, deployable encoding.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Computer Vision: Representation Learning for Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Implicit neural representation (INR) methods for video compression have recently achieved visual quality and compression ratios that are competitive with traditional pipelines. However, due to the need for per-sample network training, the encoding speeds of these methods are too slow for practical adoption. We develop a library to allow us to disentangle and review the components of methods from the NeRV family, reframing their performance in terms of not only size-quality trade-offs, but also impacts on training time. We uncover principles for effective video INR design and propose a state-of-the-art configuration of these components, Rabbit NeRV (RNeRV). When all methods are given equal training time (equivalent to 300 NeRV epochs) for 7 different UVG videos at 1080p, RNeRV achieves +1.27% PSNR on average compared to the best-performing alternative for each video in our NeRV library. We then tackle the encoding speed issue head-on by investigating the viability of hyper-networks, which predict INR weights from video inputs, to disentangle training from encoding to allow for real-time encoding. We propose masking the weights of the predicted INR during training to allow for variable, higher quality compression, resulting in 1.7% improvements to both PSNR and MS-SSIM at 0.037 bpp on the UCF-101 dataset, and we increase hyper-network parameters by 0.4% for 2.5%/2.7% improvements to PSNR/MS-SSIM with equal bpp and similar speeds. Our project website is available at https://mgwillia.github.io/vinrb/ and our code is available at https://github.com/mgwillia/vinrb.

Problem

• Improve encoding speed of implicit neural video compression
• Optimize design principles for video INR methods
• Enable real-time encoding using hyper-networks for INRs

Innovation

• Develops Rabbit NeRV for video compression
• Uses hyper-networks for real-time encoding
• Improves PSNR and MS-SSIM with masking