Superpixel-informed Implicit Neural Representation for Multi-Dimensional Data

TL;DR

This paper introduces S-INR, a superpixel-informed implicit neural representation that leverages semantic priors for improved multi-dimensional data recovery, outperforming existing INR methods.

Contribution

The paper proposes a novel superpixel-based approach for INRs, integrating attention mechanisms and shared dictionaries to exploit semantic information more effectively.

Findings

S-INR outperforms state-of-the-art INR methods in various applications.

Utilizing superpixels enhances the semantic understanding in data recovery.

The approach is effective for multi-dimensional data like images and weather data.

Abstract

Recently, implicit neural representations (INRs) have attracted increasing attention for multi-dimensional data recovery. However, INRs simply map coordinates via a multi-layer perception (MLP) to corresponding values, ignoring the inherent semantic information of the data. To leverage semantic priors from the data, we propose a novel Superpixel-informed INR (S-INR). Specifically, we suggest utilizing generalized superpixel instead of pixel as an alternative basic unit of INR for multi-dimensional data (e.g., images and weather data). The coordinates of generalized superpixels are first fed into exclusive attention-based MLPs, and then the intermediate results interact with a shared dictionary matrix. The elaborately designed modules in S-INR allow us to ingenuously exploit the semantic information within and across generalized superpixels. Extensive experiments on various applications validate the effectiveness and efficacy of our S-INR compared to state-of-the-art INR methods.