Datasets › Real Blur Dataset

Real Blur Dataset

Introduced by Jaesung Rim et al. in Real-World Blur Dataset for Learning and Benchmarking Deblurring Algorithms1 Aug 2020 archive 2025-07-28

The dataset consists of 4,738 pairs of images of 232 different scenes including reference pairs. All images were captured both in the camera raw and JPEG formats, hence generating two datasets: RealBlur-R from the raw images, and RealBlur-J from the JPEG images. Each training set consists of 3,758 image pairs, while each test set consists of 980 image pairs.

The deblurring result is first aligned to its ground truth sharp image using a homography estimated by the enhanced correlation coefficients method, and PSNR or SSIM is computed in sRGB color space.

Benchmarks archive 2025-07-28

All 6 leaderboards whose dataset resolves to this page shown (sort by any header). "First row" is the archive's own first row at snapshot, in the archive's row order; nothing here re-ranks and metric direction is not asserted.

First row (archive order)PaperCode
Deblurring RealBlur-R (trained on GoPro) AdaRevD PSNR (sRGB) 36.53 AdaRevD: Adaptive Patch Exiting Reversible Decoder... invokerer/deeprft +3 19 Compare
Deblurring RealBlur-J AdaRevD PSNR (sRGB) 33.96 AdaRevD: Adaptive Patch Exiting Reversible Decoder... invokerer/deeprft +3 17 Compare
Deblurring RealBlur-R AdaRevD PSNR (sRGB) 41.19 AdaRevD: Adaptive Patch Exiting Reversible Decoder... invokerer/deeprft +3 17 Compare
Deblurring RealBlur-J (trained on GoPro) AdaRevD PSNR (sRGB) 30.12 AdaRevD: Adaptive Patch Exiting Reversible Decoder... invokerer/deeprft +3 15 Compare
Image Deblurring RealBlur-J EVSSM PSNR 34.15 Efficient Visual State Space Model for Image Deblurring kkkls/evssm 1 Compare
Image Deblurring RealBlur-R EVSSM PSNR 41.04 Efficient Visual State Space Model for Image Deblurring kkkls/evssm 1 Compare

Papers archive 2025-07-28

28 shown of 28 papers with a leaderboard row on this dataset's benchmarks, newest first. The archive's own "papers using this dataset" list was never published, so this is the benchmark-backed subset; the archive's count for this dataset is 100. The Syntology column is from Syntology's graph (read 2026-09-24), stated per sample; it is not part of any archive number.

DateSamples run Syntology
LoFormer: Local Frequency Transformer for Image Deblurring 2 2 24 Jul 2024 not harvested
AdaRevD: Adaptive Patch Exiting Reversible Decoder Pushes the Limit of Image Deblurring 4 4 13 Jun 2024 ran 36 of 50 samples (14 unverified; 50 pointer-only for licence)
Efficient Visual State Space Model for Image Deblurring 1 2 23 May 2024 ran 2 of 2 samples (0 unverified)
Learning Enriched Features via Selective State Spaces Model for Efficient Image Deblurring 1 4 29 Mar 2024 not harvested
DeblurDiNAT: A Compact Model with Exceptional Generalization and Visual Fidelity on Unseen Domains 1 2 19 Mar 2024 not harvested
Efficient Multi-scale Network with Learnable Discrete Wavelet Transform for Blind Motion Deblurring 1 2 29 Dec 2023 ran 9 of 12 samples (3 unverified; 12 pointer-only for licence)
ID-Blau: Image Deblurring by Implicit Diffusion-based reBLurring AUgmentation 1 6 18 Dec 2023 ran 1 of 1 samples (0 unverified; 1 pointer-only for licence)
Revisiting Image Deblurring with an Efficient ConvNet 1 1 4 Feb 2023 not harvested
Self-Supervised Non-Uniform Kernel Estimation With Flow-Based Motion Prior for Blind Image Deblurring 1 1 1 Jan 2023 not harvested
Efficient Frequency Domain-based Transformers for High-Quality Image Deblurring 1 2 22 Nov 2022 ran 2 of 2 samples (0 unverified; 2 pointer-only for licence)
Stripformer: Strip Transformer for Fast Image Deblurring 2 2 10 Apr 2022 ran 5 of 8 samples (3 unverified; 8 pointer-only for licence)
MSSNet: Multi-Scale-Stage Network for Single Image Deblurring 1 4 19 Feb 2022 not harvested
MAXIM: Multi-Axis MLP for Image Processing 3 5 9 Jan 2022 ran 27 of 46 samples (19 unverified)
Intriguing Findings of Frequency Selection for Image Deblurring 5 4 23 Nov 2021 not harvested
Restormer: Efficient Transformer for High-Resolution Image Restoration 13 2 18 Nov 2021 ran 4 of 4 samples (0 unverified; 4 pointer-only for licence)
Rethinking Coarse-to-Fine Approach in Single Image Deblurring 4 1 11 Aug 2021 ran 1 of 9 samples (8 unverified; 6 pointer-only for licence)
Uformer: A General U-Shaped Transformer for Image Restoration 4 2 6 Jun 2021 ran 6 of 6 samples (0 unverified; 2 pointer-only for licence)
Multi-Stage Progressive Image Restoration 8 4 4 Feb 2021 ran 18 of 26 samples (8 unverified; 25 pointer-only for licence)
BANet: Blur-aware Attention Networks for Dynamic Scene Deblurring 1 2 19 Jan 2021 not harvested
DeblurGAN-v2: Deblurring (Orders-of-Magnitude) Faster and Better 6 4 10 Aug 2019 not harvested
Deep Stacked Hierarchical Multi-patch Network for Image Deblurring 1 1 6 Apr 2019 not harvested
Dynamic Scene Deblurring Using Spatially Variant Recurrent Neural Networks 1 1 1 Jun 2018 not harvested
Scale-recurrent Network for Deep Image Deblurring 4 4 6 Feb 2018 not harvested
DeblurGAN: Blind Motion Deblurring Using Conditional Adversarial Networks 13 2 19 Nov 2017 not harvested
Deep Multi-scale Convolutional Neural Network for Dynamic Scene Deblurring 1 1 7 Dec 2016 ran 0 of 11 samples (11 unverified)
Blind Image Deblurring Using Dark Channel Prior 0 2 1 Jun 2016 not harvested
Deblurring Low-light Images with Light Streaks 0 2 1 Jun 2014 not harvested
Unnatural L0 Sparse Representation for Natural Image Deblurring 0 1 1 Jun 2013 not harvested

Dataset loaders archive 2025-07-28

pytorch

1 loader as listed in the archive; links are outbound and not re-checked here.

Tasks archive 2025-07-28

License archive 2025-07-28

No licence recorded in the archive. Absence here is not a statement about the dataset's terms.

Modalities archive 2025-07-28

Languages archive 2025-07-28

No language tagged.

Variants archive 2025-07-28

  • RealBlur-J
  • Real Blur Dataset
  • RealBlur-R
  • RealBlur-R (trained on GoPro)
  • RealBlur-J (trained on GoPro)

5 variant names, as the archive lists them.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections