MUJICA: Reforming SISR Models for PBR Material Super-Resolution via Cross-Map Attention

TL;DR

This paper introduces MUJICA, a novel method that adapts pre-trained SISR models for PBR material super-resolution by using cross-map attention to enhance feature fusion and consistency across multiple texture maps.

Contribution

MUJICA is a flexible adapter that reformulates existing SISR models for PBR materials, improving super-resolution quality and cross-map consistency with minimal retraining.

Findings

MUJICA improves PSNR, SSIM, and LPIPS scores on PBR datasets.

It enables efficient training with limited resources.

MUJICA achieves state-of-the-art performance in PBR material super-resolution.

Abstract

Physically Based Rendering (PBR) materials are typically characterized by multiple 2D texture maps such as basecolor, normal, metallic, and roughness which encode spatially-varying bi-directional reflectance distribution function (SVBRDF) parameters to model surface reflectance properties and microfacet interactions. Upscaling SVBRDF material is valuable for modern 3D graphics applications. However, existing Single Image Super-Resolution (SISR) methods struggle with cross-map inconsistency, inadequate modeling of modality-specific features, and limited generalization due to data distribution shifts. In this work, we propose Multi-modal Upscaling Joint Inference via Cross-map Attention (MUJICA), a flexible adapter that reforms pre-trained Swin-transformer-based SISR models for PBR material super-resolution. MUJICA is seamlessly attached after the pre-trained and frozen SISR backbone. It leverages cross-map attention to fuse features while preserving remarkable reconstruction ability of the pre-trained SISR model. Applied to SISR models such as SwinIR, DRCT, and HMANet, MUJICA improves PSNR, SSIM, and LPIPS scores while preserving cross-map consistency. Experiments demonstrate that MUJICA enables efficient training even with limited resources and delivers state-of-the-art performance on PBR material datasets.