NTIRE-Haze: Real Haze-Machine Dehazing Benchmarks (Unofficial Composite Mirror)

Unofficial composite redistribution of the four NTIRE real-haze dehazing benchmarks, I-Haze, O-Haze, Dense-Haze, and NH-Haze (Ancuti et al., 2018-2020), packaged for direct use with ClearView's dataset pipeline. This is ClearView's first dehazing dataset, a degradation category distinct from rain streaks, raindrops, or snow.

Disclaimer

This repository is not an official release of I-Haze, O-Haze, Dense-Haze, or NH-Haze.

All four sets were created by Codruta O. Ancuti, Cosmin Ancuti, and collaborators (full author lists per constituent below). This repository does not claim ownership of any images or metadata, and attributes each constituent set to its original creators. We describe them as "creators" rather than "copyright holders" deliberately, copyright in academic datasets can rest with an author's university, a funding body, or otherwise, under arrangements we have no way to verify from the outside.

On licensing, please read before relying on this mirror. We looked for a formal license for each of the four constituent sets individually, and found none in any case:

Constituent	Original paper	License found?
I-Haze	Ancuti et al., ACIVS 2018	None
O-Haze	Ancuti et al., CVPRW 2018	None
Dense-Haze	Ancuti et al., ICIP 2019	None
NH-Haze	Ancuti et al., CVPRW 2020	None

This is the same situation as this collection's Rain13K mirror (a five-way composite with an identical per-constituent license search), not a stated restriction like a dataset that explicitly says "research purposes only" (which we treat differently and do not mirror, see this collection's decision on GT-RAIN for that case). In the absence of an explicit license, we have chosen to preserve these datasets for research reproducibility while providing full attribution and making the mirror removable on request, this is a decision we're making under that uncertainty, not a claim that we've established a legal right to redistribute it. Concretely:

• Full attribution and citation to every constituent set's original creators (below).
• No claim of any rights beyond hosting a copy.
• If a copyright holder, or an authorized representative, for any constituent set requests removal, we will comply promptly, for that specific constituent, or the whole bundle if requested.

If you need certainty about redistribution rights (e.g. for commercial use), contact the relevant dataset's creators or their institution directly rather than relying on this mirror.

---

Dataset Overview

All four NTIRE haze sets share the same real-world capture methodology: professional haze machines generate physical fog/haze in a real scene, and a hazy/haze-free pair is captured under otherwise identical lighting and camera position. This makes them real haze, not synthetically composited, the same reason SPA-Data and RealRain-1k are used as real-world checks against synthetic rain in this collection.

Constituent	Pairs	Domain	Notes
I-Haze	30	Indoor	ACIVS 2018.
O-Haze	45	Outdoor	CVPRW 2018, companion to I-Haze.
Dense-Haze	55	Outdoor	ICIP 2019, very dense fog, harder than O-Haze.
NH-Haze	55	Outdoor	CVPRW 2020, non-homogeneous (uneven) haze, closer to real patchy fog.
Total	185

Note: several secondary sources cite different pair counts (e.g. "25/35 I-Haze", "33 Dense-Haze"), this card's counts are verified directly against the actual downloaded, extracted files, not repeated from secondhand citations.

---

Changes from the Official Release

No images added, removed, or modified in any constituent set. Two repackaging-only changes:

• Folder rename. I-Haze and O-Haze's official archives ship an outer wrapper folder literally named # I-HAZY NTIRE 2018 / # O-HAZY NTIRE 2018 (with a leading # character), which is awkward for tooling and URLs. This mirror drops that wrapper and uses a plain I-HAZE/ / O-HAZE/ top-level folder instead, the GT/hazy subfolders and every filename inside are untouched.
• Composite bundling. The four sets are hosted together in one repository (mirroring this collection's Rain13K precedent) rather than four separate repos, since they share a license situation, capture methodology, and task.

---

Dataset Structure

ntire-haze/
├── README.md
├── banner.jpg
├── I-HAZE/
│   ├── GT/      # {id}_indoor_GT.jpg    (30 images)
│   └── hazy/    # {id}_indoor_hazy.jpg  (30 images)
├── O-HAZE/
│   ├── GT/      # {id}_outdoor_GT.ext   (45 images, mixed .jpg/.JPG)
│   └── hazy/    # {id}_outdoor_hazy.ext (45 images, mixed .jpg/.JPG)
├── Dense-Haze/
│   ├── GT/      # {id}_GT.png    (55 images)
│   └── hazy/    # {id}_hazy.png  (55 images)
└── NH-HAZE/     # flat, no GT/hazy subfolders
    ├── {id}_GT.png    (55 images)
    └── {id}_hazy.png  (55 images)

Filenames mark image type with a _GT/_hazy suffix, sometimes with an extra domain tag in between (I-Haze/O-Haze's _indoor/_outdoor), this is why ClearView ships a dedicated NTIREHazeDataset parser: it normalizes across the GT/hazy-subfolder layout (I-Haze/O-Haze/Dense-Haze) and the flat layout (NH-Haze) transparently, matching by stripping the suffix marker case-insensitively (handles O-Haze's mixed .jpg/.JPG extensions too).

---

Usage with ClearView

This dataset is designed to be used directly with ClearView, an open-source PyTorch framework for image restoration.

from huggingface_hub import snapshot_download
from clearview.data import NTIREHazeDataset, get_train_transforms

data_dir = snapshot_download(repo_id="dronefreak/NTIRE-Haze", repo_type="dataset")

ihaze = NTIREHazeDataset(f"{data_dir}/I-HAZE", transform=get_train_transforms(crop_size=(256, 256)))
ohaze = NTIREHazeDataset(f"{data_dir}/O-HAZE")
densehaze = NTIREHazeDataset(f"{data_dir}/Dense-Haze")
nhhaze = NTIREHazeDataset(f"{data_dir}/NH-HAZE")  # flat layout, same call

hazy, clean = ihaze[0]

Note: NTIREHazeDataset is not yet wired into the clearview-train CLI's --dataset-type choices (Python API only for now).

---

Dataset Sources

I-Haze

I-HAZE: a dehazing benchmark with real hazy and haze-free indoor images

Codruta O. Ancuti, Cosmin Ancuti, Radu Timofte, Christophe De Vleeschouwer

Advanced Concepts for Intelligent Vision Systems (ACIVS), 2018.

• arXiv: https://arxiv.org/abs/1804.05091

O-Haze

O-HAZE: a dehazing benchmark with real hazy and haze-free outdoor images

Codruta O. Ancuti, Cosmin Ancuti, Radu Timofte, Christophe De Vleeschouwer

IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2018.

• arXiv: https://arxiv.org/abs/1804.05101

Dense-Haze

Dense-Haze: A Benchmark for Image Dehazing with Dense-Haze and Haze-Free Images

Codruta O. Ancuti, Cosmin Ancuti, Mateu Sbert, Radu Timofte

IEEE International Conference on Image Processing (ICIP), 2019, pp. 1014-1018.

NH-Haze

NH-HAZE: An Image Dehazing Benchmark with Non-Homogeneous Hazy and Haze-Free Images

Codruta O. Ancuti, Cosmin Ancuti, Radu Timofte

IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2020.

• arXiv: https://arxiv.org/abs/2005.03560

Official Resources

• I-Haze: https://data.vision.ee.ethz.ch/cvl/ntire18/i-haze/
• O-Haze: https://data.vision.ee.ethz.ch/cvl/ntire18/o-haze/
• Dense-Haze: https://data.vision.ee.ethz.ch/cvl/ntire19/dense-haze/
• NH-Haze: https://data.vision.ee.ethz.ch/cvl/ntire20/nh-haze/

---

Attribution

All credit for these datasets belongs entirely to the original authors: Codruta O. Ancuti, Cosmin Ancuti, Radu Timofte, Christophe De Vleeschouwer, and Mateu Sbert (across the four constituent papers, author list varies per paper as listed above).

If you use this dataset in your research, please cite the specific constituent paper(s) you use, listed below.

---

License

No formal license was located for any of the four constituent sets, see the Disclaimer above for our reasoning for mirroring them anyway under that uncertainty (full attribution, reproducibility purpose, removable on request). This repository makes no claim to any rights over the dataset content beyond hosting a copy.

---

Citation

If you use this dataset, please cite the relevant constituent paper(s):

@inproceedings{ancuti2018ihaze,
  author = {Ancuti, Codruta O. and Ancuti, Cosmin and Timofte, Radu and De Vleeschouwer, Christophe},
  title = {I-HAZE: A Dehazing Benchmark with Real Hazy and Haze-Free Indoor Images},
  booktitle = {Advanced Concepts for Intelligent Vision Systems (ACIVS)},
  pages = {620--631},
  year = {2018}
}

@inproceedings{ancuti2018ohaze,
  author = {Ancuti, Codruta O. and Ancuti, Cosmin and Timofte, Radu and De Vleeschouwer, Christophe},
  title = {O-HAZE: A Dehazing Benchmark with Real Hazy and Haze-Free Outdoor Images},
  booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)},
  year = {2018}
}

@inproceedings{ancuti2019densehaze,
  author = {Ancuti, Codruta O. and Ancuti, Cosmin and Sbert, Mateu and Timofte, Radu},
  title = {Dense-Haze: A Benchmark for Image Dehazing with Dense-Haze and Haze-Free Images},
  booktitle = {IEEE International Conference on Image Processing (ICIP)},
  pages = {1014--1018},
  year = {2019}
}

@inproceedings{ancuti2020nhhaze,
  author = {Ancuti, Codruta O. and Ancuti, Cosmin and Timofte, Radu},
  title = {NH-HAZE: An Image Dehazing Benchmark with Non-Homogeneous Hazy and Haze-Free Images},
  booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)},
  month = {June},
  year = {2020}
}

---

Acknowledgements

We sincerely thank Codruta O. Ancuti, Cosmin Ancuti, Radu Timofte, Christophe De Vleeschouwer, and Mateu Sbert for creating and publicly releasing these foundational real-haze dehazing benchmarks.