FocalSR: Revisiting image super-resolution transformers with fourier-transform cross attention layers for remote sensing image enhancement
Bibliographic record
Abstract
Transformer architecture has attained noteworthy performance achievements in recent image super-resolution research. However, current transformer-based methods still expose limitations in fully harnessing domain-specific information within images, particularly when applied to broader-scale remote sensing images that contain diverse landscape objects on one scene. Remote sensing images have relatively lower resolution compared to the common super-resolution training dataset and each landscape object covers a small area on the image. These natures of remote sensing images significantly reduced the attention pixels for image restoration in existing transformer-based methods. To address this challenge and enhance domain-specific multi-object image reconstruction, we introduce FocalSR, a Transformer model featuring FOurier-transform Cross Attention Layers for Super-Resolution. Drawing inspiration from state-of-the-art Transformer models like Hybrid Attention Transformer (HAT), FocalSR incorporates channel-focused and window-centric self-attention mechanisms. By integrating Fast Fourier Convolution into the cross-attention layer, FocalSR extends its capacity to capture image-wide information and intricate details in low-resolution images. Through unified task pretraining during model development, we validate the efficacy of these enhancements through extensive testing, resulting in substantial performance improvements. Notably, our experiments showcase FocalSR's superior performance in remote sensing datasets, demonstrating a notable 1 dB enhancement in the PSNR metric compared to other state-of-the-art methods. Additionally, significant improvements are observed in challenging scenarios such as pattern restoration and vegetation detail preservation, underscoring the transformative potential of FocalSR in advancing image processing and domain-specific vision tasks. • Design a novel super-resolution network for remote sensing image enhancement. • Incorporate fast Fourier convolution to improve the model performance. • Demonstrate the potentials to reveal finer spatial patterns and canopy details.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".