FINER: Flexible spectral-bias tuning in Implicit NEural Representation by Variable-periodic Activation Functions

TL;DR

FINER introduces variable-periodic activation functions with tunable biases in implicit neural representations, enabling flexible frequency support and improved performance across various signal representation tasks.

Contribution

The paper proposes FINER, a novel method that allows flexible spectral bias tuning in INRs through bias initialization, enhancing their ability to represent complex signals.

Findings

Outperforms existing INRs in 2D image fitting

Achieves better 3D signed distance field representation

Improves 5D neural radiance fields optimization

Abstract

Implicit Neural Representation (INR), which utilizes a neural network to map coordinate inputs to corresponding attributes, is causing a revolution in the field of signal processing. However, current INR techniques suffer from a restricted capability to tune their supported frequency set, resulting in imperfect performance when representing complex signals with multiple frequencies. We have identified that this frequency-related problem can be greatly alleviated by introducing variable-periodic activation functions, for which we propose FINER. By initializing the bias of the neural network within different ranges, sub-functions with various frequencies in the variable-periodic function are selected for activation. Consequently, the supported frequency set of FINER can be flexibly tuned, leading to improved performance in signal representation. We demonstrate the capabilities of FINER in the contexts of 2D image fitting, 3D signed distance field representation, and 5D neural radiance fields optimization, and we show that it outperforms existing INRs.