Subtractive Modulative Network with Learnable Periodic Activations

TL;DR

The paper introduces the Subtractive Modulative Network (SMN), a parameter-efficient neural architecture inspired by classical synthesis, that excels in image reconstruction and 3D view synthesis through learnable periodic activations and modulative modules.

Contribution

It presents a novel INR architecture with learnable periodic activations and modulative masks, combining theoretical analysis with empirical validation for improved efficiency and accuracy.

Findings

Achieves PSNR of 40+ dB on image datasets

Outperforms state-of-the-art methods in reconstruction accuracy

Shows consistent advantages in 3D NeRF view synthesis

Abstract

We propose the Subtractive Modulative Network (SMN), a novel, parameter-efficient Implicit Neural Representation (INR) architecture inspired by classical subtractive synthesis. The SMN is designed as a principled signal processing pipeline, featuring a learnable periodic activation layer (Oscillator) that generates a multi-frequency basis, and a series of modulative mask modules (Filters) that actively generate high-order harmonics. We provide both theoretical analysis and empirical validation for our design. Our SMN achieves a PSNR of $40+$ dB on two image datasets, comparing favorably against state-of-the-art methods in terms of both reconstruction accuracy and parameter efficiency. Furthermore, consistent advantage is observed on the challenging 3D NeRF novel view synthesis task. Supplementary materials are available at https://inrainbws.github.io/smn/.