mlsanityData to 14 Oct 2026

Winner of CVPR2026 NTIRE Challenge on Image Shadow Removal: Semantic and Geometric Guidance for Shadow Removal via Cascaded Refinement

arXiv 2604.16177 ↗First posted 17 Apr 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

We present a three-stage progressive shadow-removal pipeline for the CVPR2026 NTIRE WSRD+ challenge. Built on OmniSR, our method treats deshadowing as iterative direct refinement, where later stages correct residual artefacts left by earlier predictions. The model combines RGB appearance with frozen DINOv2 semantic guidance and geometric cues from monocular depth and surface normals, reused across all stages. To stabilise multi-stage optimisation, we introduce a contraction-constrained objective that encourages non-increasing reconstruction error across the cascade. A staged training pipeline transfers from earlier WSRD pretraining to WSRD+ supervision and final WSRD+ 2026 adaptation with cosine-annealed checkpoint ensembling. On the official WSRD+ 2026 hidden test set, our final ensemble achieved 26.680 PSNR, 0.8740 SSIM, 0.0578 LPIPS, and 26.135 FID, ranked first overall, and won the NTIRE 2026 Image Shadow Removal Challenge. The strong performance of the proposed model is further validated on the ISTD+ and UAV-SC+ datasets.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 23 of its 41 references that are in the map. First posted on arXiv 17 Apr 2026.

Builds on

23 of 41 references.

References · 23

  1. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,177
  2. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,974
  3. 2024Depth Anything V2cited by 1,422
  4. 2023Rethinking FID: Towards a Better Evaluation Metric for Image Generationcited by 310
  5. 2016The Little Engine that Could: Regularization by Denoising (RED)cited by 307
  6. 2019Plug-and-Play Methods Provably Converge with Properly Trained Denoiserscited by 198
  7. 2017Stacked Conditional Generative Adversarial Networks for Jointly Learning Shadow Detection and Shadow Removalcited by 161
  8. 2018Regularization by Denoising: Clarifications and New Interpretationscited by 103
  9. 2019Shadow Removal via Shadow Image Decompositioncited by 98
  10. 2019Mask-ShadowGAN: Learning to Remove Shadows from Unpaired Datacited by 92
  11. 2023Refusion: Enabling Large-Size Realistic Image Restoration with Latent-Space Diffusion Modelscited by 86
  12. 2021Auto-Exposure Fusion for Single-Image Shadow Removalcited by 79
11 more
  1. 2022DC-ShadowNet: Single-Image Hard and Soft Shadow Removal Using Unsupervised Domain-Classifier Guided Networkcited by 69
  2. 2021From Shadow Generation to Shadow Removalcited by 55
  3. 2020Shadow Removal by a Lightness-Guided Network with Training on Unpaired Datacited by 38
  4. 2022DeS3: Adaptive Attention-driven Self and Soft Shadow Removal using ViT Similaritycited by 38
  5. 2023ShadowFormer: Global Context Helps Image Shadow Removalcited by 35
  6. 2017Improving Shadow Suppression for Illumination Robust Face Recognitioncited by 29
  7. 2023Latent Feature-Guided Diffusion Models for Shadow Removalcited by 29
  8. 2021Learning from Synthetic Shadows for Shadow Detection and Removalcited by 27
  9. 2024OmniSR: Shadow Removal under Direct and Indirect Lightingcited by 19
  10. 2024Detail-Preserving Latent Diffusion for Stable Shadow Removalcited by 13
  11. 2024Regional Attention for Shadow Removalcited by 12

Built on by

Not cited yet.