Predicting the Encoding Error of SIRENs

TL;DR

This paper introduces a predictive method for estimating the encoding error of SIREN neural networks, enabling faster assessment of signal fidelity without extensive training, and offers insights into SIREN behavior and performance comparisons.

Contribution

The paper presents a novel regression approach trained on a large dataset to predict SIREN encoding error based on hyperparameters and signals, reducing computational costs.

Findings

Predictive model accurately estimates SIREN encoding error in milliseconds.

Narrow SIRENs exhibit high variability in encoding error.

SIREN performance shows parallels with JPEG compression results.

Abstract

Implicit Neural Representations (INRs), which encode signals such as images, videos, and 3D shapes in the weights of neural networks, are becoming increasingly popular. Among their many applications is signal compression, for which there is great interest in achieving the highest possible fidelity to the original signal subject to constraints such as neural network size, training (encoding) and inference (decoding) time. But training INRs can be a computationally expensive process, making it challenging to determine the best possible tradeoff under such constraints. Towards this goal, we present a method which predicts the encoding error that a popular INR network (SIREN) will reach, given its network hyperparameters and the signal to encode. This method is trained on a unique dataset of 300,000 SIRENs, trained across a variety of images and hyperparameters. (Dataset available here: https://huggingface.co/datasets/predict-SIREN-PSNR/COIN-collection.) Our predictive method demonstrates the feasibility of this regression problem, and allows users to anticipate the encoding error that a SIREN network will reach in milliseconds instead of minutes or longer. We also provide insights into the behavior of SIREN networks, such as why narrow SIRENs can have very high random variation in encoding error, and how the performance of SIRENs relates to JPEG compression.