mlsanityData to 14 Oct 2026

Efficient Perceptual Image Super Resolution: AIM 2025 Study and Benchmark

arXiv 2510.12765 ↗First posted 14 Oct 2025cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

This paper presents a comprehensive study and benchmark on Efficient Perceptual Super-Resolution (EPSR). While significant progress has been made in efficient PSNR-oriented super resolution, approaches focusing on perceptual quality metrics remain relatively inefficient. Motivated by this gap, we aim to replicate or improve the perceptual results of Real-ESRGAN while meeting strict efficiency constraints: a maximum of 5M parameters and 2000 GFLOPs, calculated for an input size of 960x540 pixels. The proposed solutions were evaluated on a novel dataset consisting of 500 test images of 4K resolution, each degraded using multiple degradation types, without providing the original high-quality counterparts. This design aims to reflect realistic deployment conditions and serves as a diverse and challenging benchmark. The top-performing approach manages to outperform Real-ESRGAN across all benchmark datasets, demonstrating the potential of efficient methods in the perceptual domain. This paper establishes the modern baselines for efficient perceptual super resolution.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 32 of its 50 references that are in the map. First posted on arXiv 14 Oct 2025.

Builds on

32 of 50 references.

References · 32

  1. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  2. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  3. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,177
  4. 2016Perceptual Losses for Real-Time Style Transfer and Super-Resolutioncited by 2,694
  5. 2016Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Networkcited by 1,908
  6. 2017Enhanced Deep Residual Networks for Single Image Super-Resolutioncited by 1,552
  7. 2021Image Super-Resolution via Iterative Refinementcited by 1,511
  8. 2022Exploring CLIP for Assessing the Look and Feel of Imagescited by 991
  9. 2021Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Datacited by 970
  10. 2017Learning Efficient Convolutional Networks through Network Slimmingcited by 676
  11. 2017The Perception-Distortion Tradeoffcited by 587
  12. 2021Designing a Practical Degradation Model for Deep Blind Image Super-Resolutioncited by 529
20 more
  1. 2022MANIQA: Multi-dimension Attention Network for No-Reference Image Quality Assessmentcited by 522
  2. 2019Toward Real-World Single Image Super-Resolution: A New Benchmark and A New Modelcited by 403
  3. 2018Recovering Realistic Texture in Image Super-resolution by Deep Spatial Feature Transformcited by 371
  4. 2020Real-Time Intermediate Flow Estimation for Video Frame Interpolationcited by 213
  5. 2019ACNet: Strengthening the Kernel Skeletons for Powerful CNN via Asymmetric Convolution Blockscited by 178
  6. 2020PIPAL: a Large-Scale Image Quality Assessment Dataset for Perceptual Image Restorationcited by 178
  7. 2018The 2018 PIRM Challenge on Perceptual Image Super-resolutioncited by 156
  8. 2022Details or Artifacts: A Locally Discriminative Learning Approach to Realistic Image Super-Resolutioncited by 114
  9. 2021Diverse Branch Block: Building a Convolution as an Inception-like Unitcited by 87
  10. 2022Swin2SR: SwinV2 Transformer for Compressed Image Super-Resolution and Restorationcited by 73
  11. 2023Spatially-Adaptive Feature Modulation for Efficient Image Super-Resolutioncited by 57
  12. 2024The Ninth NTIRE 2024 Efficient Super-Resolution Challenge Reportcited by 57
  13. 2022NTIRE 2022 Challenge on Efficient Super-Resolution: Methods and Resultscited by 40
  14. 2023Swift Parameter-free Attention Network for Efficient Super-Resolutioncited by 24
  15. 2022Edge-enhanced Feature Distillation Network for Efficient Super-Resolutioncited by 18
  16. 2025AIM 2025 Low-light RAW Video Denoising Challenge: Dataset, Methods and Resultscited by 14
  17. 2022Gradient Variance Loss for Structure-Enhanced Image Super-Resolutioncited by 12
  18. 2025AIM 2025 challenge on Inverse Tone Mapping Report: Methods and Resultscited by 8
  19. 2025AIM 2025 Rip Current Segmentation (RipSeg) Challenge Reportcited by 7
  20. 2024Auto-Encoded Supervision for Perceptual Image Super-Resolutioncited by 4

Built on by

Not cited yet.