Saved in:
Bibliographic Details
Main Authors: Zeraatkar, Ehsan, Faroughi, Salah A, Tešić, Jelena
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2502.12427
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914335547195392
author Zeraatkar, Ehsan
Faroughi, Salah A
Tešić, Jelena
author_facet Zeraatkar, Ehsan
Faroughi, Salah A
Tešić, Jelena
contents Super-resolution can play an essential role in enhancing the spatial fidelity of Earth System Model outputs, allowing fine-scale structures highly beneficial to climate science to be recovered from coarse simulations. However, traditional deep super-resolution methods, including convolutional and transformer based models, tend to exhibit spectral bias, reconstructing low-frequency content more readily than valuable high-frequency details. In this work, we introduce ViSIR and ViFOR, two frequency-aware frameworks. ViSIR stands for the Vision Transformer-Tuned Sinusoidal Implicit Representation. ViSIR combines vision transformers with sinusoidal activations to mitigate spectral bias. ViFOR stands for the Vision Transformer Fourier Representation Network. ViFOR integrates explicit Fourier based filtering for independent low- and high-frequency learning. Evaluated on the E3SM-HR Earth system dataset across surface temperature, shortwave, and longwave fluxes, these models outperform leading Convolutional NN, Generative Networks, and vanilla transformer baselines, with ViFOR demonstrating up to 2.6~dB improvements in Peak Signal to Noise Ratio and higher Structural Similarity.
format Preprint
id arxiv_https___arxiv_org_abs_2502_12427
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Frequency-Aware Vision Transformers for High-Fidelity Super-Resolution of Earth System Models
Zeraatkar, Ehsan
Faroughi, Salah A
Tešić, Jelena
Computer Vision and Pattern Recognition
Super-resolution can play an essential role in enhancing the spatial fidelity of Earth System Model outputs, allowing fine-scale structures highly beneficial to climate science to be recovered from coarse simulations. However, traditional deep super-resolution methods, including convolutional and transformer based models, tend to exhibit spectral bias, reconstructing low-frequency content more readily than valuable high-frequency details. In this work, we introduce ViSIR and ViFOR, two frequency-aware frameworks. ViSIR stands for the Vision Transformer-Tuned Sinusoidal Implicit Representation. ViSIR combines vision transformers with sinusoidal activations to mitigate spectral bias. ViFOR stands for the Vision Transformer Fourier Representation Network. ViFOR integrates explicit Fourier based filtering for independent low- and high-frequency learning. Evaluated on the E3SM-HR Earth system dataset across surface temperature, shortwave, and longwave fluxes, these models outperform leading Convolutional NN, Generative Networks, and vanilla transformer baselines, with ViFOR demonstrating up to 2.6~dB improvements in Peak Signal to Noise Ratio and higher Structural Similarity.
title Frequency-Aware Vision Transformers for High-Fidelity Super-Resolution of Earth System Models
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2502.12427