VoDaSuRe: A Large-Scale Dataset Revealing Domain Shift in Volumetric Super-Resolution

TL;DR

This paper introduces VoDaSuRe, a large-scale dataset of paired high- and low-resolution volumetric scans, revealing that current super-resolution models trained on downsampled data do not perform well on real low-resolution scans, highlighting a gap in the field.

Contribution

The authors present VoDaSuRe, the first large-scale dataset of real paired volumetric scans, and demonstrate the discrepancy between models trained on downsampled versus real low-resolution data.

Findings

Models trained on downsampled data produce sharper predictions.

Models trained on real low-resolution scans preserve structure but are less accurate.

Current SR methods are overstated when applied to real data.

Abstract

Recent advances in volumetric super-resolution (SR) have demonstrated strong performance in medical and scientific imaging, with transformer- and CNN-based approaches achieving impressive results even at extreme scaling factors. In this work, we show that much of this performance stems from training on downsampled data rather than real low-resolution scans. This reliance on downsampling is partly driven by the scarcity of paired high- and low-resolution 3D datasets. To address this, we introduce VoDaSuRe, a large-scale volumetric dataset containing paired high- and low-resolution scans. When training models on VoDaSuRe, we reveal a significant discrepancy: SR models trained on downsampled data produce substantially sharper predictions than those trained on real low-resolution scans, which smooth fine structures. Conversely, applying models trained on downsampled data to real scans preserves more structure but is inaccurate. Our findings suggest that current SR methods are overstated - when applied to real data, they do not recover structures lost in low-resolution scans and instead predict a smoothed average. We argue that progress in deep learning-based volumetric SR requires datasets with paired real scans of high complexity, such as VoDaSuRe. Our dataset and code are publicly available through: https://augusthoeg.github.io/VoDaSuRe/