mlsanityData to 14 Oct 2026

ESSR: An 8K@30FPS Super-Resolution Accelerator With Edge Selective Network

arXiv 2503.20245 ↗First posted 26 Mar 2025cs.AR cs.AI cs.MMcited by 0

Not on the map. Why?

Abstract (arXiv)

Deep learning-based super-resolution (SR) is challenging to implement in resource-constrained edge devices for resolutions beyond full HD due to its high computational complexity and memory bandwidth requirements. This paper introduces an 8K@30FPS SR accelerator with edge-selective dynamic input processing. Dynamic processing chooses the appropriate subnets for different patches based on simple input edge criteria, achieving a 50% MAC reduction with only a 0.1dB PSNR decrease. The quality of reconstruction images is guaranteed and maximized its potential with textitresource adaptive model switching even under resource constraints. In conjunction with hardware-specific refinements, the model size is reduced by 84% to 51K, but with a decrease of less than 0.6dB PSNR. Additionally, to support dynamic processing with high utilization, this design incorporates a textitconfigurable group of layer mapping that synergizes with the textitstructure-friendly fusion block, resulting in 77% hardware utilization and up to 79% reduction in feature SRAM access. The implementation, using the TSMC 28nm process, can achieve 8K@30FPS throughput at 800MHz with a gate count of 2749K, 0.2075W power consumption, and 4797Mpixels/J energy efficiency, exceeding previous work.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 21 of its 28 references that are in the map. First posted on arXiv 26 Mar 2025.

Builds on

21 of 28 references.

References · 21

  1. 2017MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applicationscited by 3,845
  2. 2016Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Networkcited by 1,908
  3. 2017Enhanced Deep Residual Networks for Single Image Super-Resolutioncited by 1,552
  4. 2018ESRGAN: Enhanced Super-Resolution Generative Adversarial Networkscited by 1,254
  5. 2021Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Datacited by 970
  6. 2017Deep Laplacian Pyramid Networks for Fast and Accurate Super-Resolutioncited by 453
  7. 2016Accelerating the Super-Resolution Convolutional Neural Networkcited by 394
  8. 2017Fast and Accurate Image Super-Resolution with Deep Laplacian Pyramid Networkscited by 138
  9. 2020Residual Feature Distillation Network for Lightweight Image Super-Resolutioncited by 122
  10. 2022Details or Artifacts: A Locally Discriminative Learning Approach to Realistic Image Super-Resolutioncited by 114
  11. 2021ClassSR: A General Framework to Accelerate Super-Resolution Networks by Data Characteristiccited by 81
  12. 2022Residual Local Feature Network for Efficient Super-Resolutioncited by 53
9 more
  1. 2020PAMS: Quantized Super-Resolution via Parameterized Max Scalecited by 48
  2. 2022NTIRE 2022 Challenge on Efficient Super-Resolution: Methods and Resultscited by 40
  3. 2022CADyQ: Content-Aware Dynamic Quantization for Image Super-Resolutioncited by 27
  4. 2022ARM: Any-Time Super-Resolution Methodcited by 19
  5. 2021Asymmetric CNN for image super-resolutioncited by 14
  6. 2022Generative Adversarial Super-Resolution at the Edge with Knowledge Distillationcited by 9
  7. 2019eCNN: A Block-Based and Highly-Parallel CNN Accelerator for Edge Inferencecited by 6
  8. 2022BSRA: Block-based Super Resolution Accelerator with Hardware Efficient Pixel Attentioncited by 2
  9. 2023ACNPU: A 4.75TOPS/W 1080P@30FPS Super Resolution Accelerator with Decoupled Asymmetric Convolutioncited by 1

Built on by

Not cited yet.