pith. sign in

arxiv: 2605.15906 · v1 · pith:JZJMDJ3Hnew · submitted 2026-05-15 · 💻 cs.CV

A Causally Grounded Taxonomy for Image Degradation Robustness Evaluation

Stefan Becker , Simon Weiss , Wolfgang H\"ubner , Michael Arens This is my paper

Pith reviewed 2026-05-20 19:03 UTC · model grok-4.3

classification 💻 cs.CV
keywords image degradationrobustness evaluationtaxonomycausal frameworkperceptual effectseverity measurementobject detectionimaging pipeline
0
0 comments X

The pith

A dual-axis taxonomy groups image degradations by causal source in the imaging pipeline and by perceptual effect, with a measurement layer that quantifies severity comparably across backends.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes a framework that overlays structure on existing image degradation benchmarks rather than creating new ones. Degradations are placed at the intersection of a causal-source axis with four pipeline stages and a perceptual-effect axis. A separate severity layer then applies standard full-reference metrics to every native level in any given implementation. This approach lets researchers compare robustness results across synthetic corruption suites, perceptual quality studies, and physically grounded camera analyses without forcing incompatible parameterizations into a single scale. The framework is illustrated by aligning degradations for an object-detection benchmark on COCO images.

Core claim

Each degradation is described along two orthogonal axes: its dominant causal source (environment, sensor/optics, ISP/renderer/codec, or transfer/system) and its resulting perceptual effect. This dual-axis abstraction produces a compact taxonomy that covers algorithmic corruptions, perceptual distortions, and physically motivated imaging artifacts. Severity is made observable and comparable by computing PSNR, SSIM, and LPIPS values for every native severity level supplied by an existing backend, leaving those native parameterizations unchanged.

What carries the argument

Dual orthogonal axes (causal source in the imaging pipeline and perceptual effect) together with a lightweight severity measurement layer that applies full-reference image quality metrics to native parameter settings.

If this is right

  • Robustness scores obtained on synthetic corruption benchmarks become directly comparable with scores from perceptual-quality or camera-failure studies.
  • The same native degradation implementations can be reused while their effective strengths are reported on a common numerical scale.
  • Gaps in coverage of the taxonomy become visible, guiding the addition of new test cases without redefining existing parameter spaces.
  • Object-detector evaluations can be conducted under a single aligned set of conditions that mixes algorithmic, perceptual, and physical degradations.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The taxonomy could serve as a checklist when designing new robustness datasets, ensuring each quadrant of the two axes receives representative test cases.
  • If the severity layer correlates with downstream task error rates, it may allow calibration of benchmark difficulty across entirely different vision tasks.
  • Extending the causal axis to video pipelines or multi-sensor systems would require only the addition of new source labels while preserving the existing perceptual axis.

Load-bearing premise

The four causal-source categories are exhaustive, mutually exclusive, and remain orthogonal to the perceptual-effect axis.

What would settle it

A collection of real or synthetic degradations that cannot be assigned to exactly one causal-source category without overlap or omission, or for which the chosen quality metrics produce inconsistent ordering of severity levels across different backends.

Figures

Figures reproduced from arXiv: 2605.15906 by Michael Arens, Simon Weiss, Stefan Becker, Wolfgang H\"ubner.

Figure 1
Figure 1. Figure 1: Backend ambiguity for identically named degradations. Two implementations labeled ’Gaussian Blur’ are applied to [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Schematic overview of the proposed image degradation taxonomy. Degradations are described by a causal source ( [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Severity calibration for ’Gaussian Blur’ from [ [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Relative mAP@0.5:0.95 drop of YOLOv10-m under increasing native severity. Measured severity semantics help explain [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Relative detection performance drop of YOLOv9-m [ [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Absolute detection performance of YOLOv9-m [ [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Relative detection performance drop of YOLOv10-m [ [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Absolute detection performance (mAP@0.5:0.95) of YOLOv10-m [ [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Relative detection performance drop of YOLO11-m [ [PITH_FULL_IMAGE:figures/full_fig_p010_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Absolute detection performance of YOLO11-m [ [PITH_FULL_IMAGE:figures/full_fig_p011_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Relative detection performance drop of RT-DETR-l [ [PITH_FULL_IMAGE:figures/full_fig_p011_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Absolute detection performance of RT-DETR-l [ [PITH_FULL_IMAGE:figures/full_fig_p011_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Extension of canonical severity levels for ’Gaussian Blur’ using fixed-step extrapolation in distortion space. Red [PITH_FULL_IMAGE:figures/full_fig_p013_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Visual examples of degradations from [9] at native severity level 3 on an example COCO image [4]. SUPPLEMENTARY EVALUATION: DEGRADATION FUNCTIONS Visual examples of the degradations applied to a representative COCO image are shown in Figures 16, 15, and 14. These galleries provide qualitative context for the degradation families evaluated in the main paper and supplementary detector comparisons [PITH_FUL… view at source ↗
Figure 15
Figure 15. Figure 15: Visual examples of the 24 degradations from [ [PITH_FULL_IMAGE:figures/full_fig_p016_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Visualizations of the 19 degradation function in [ [PITH_FULL_IMAGE:figures/full_fig_p017_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Visual examples of the ’Elastic Transform’ degradation from [ [PITH_FULL_IMAGE:figures/full_fig_p017_17.png] view at source ↗
read the original abstract

Image degradations can occur during acquisition, processing, and transmission, altering visual appearance and affecting downstream vision tasks. They are studied in several communities, including synthetic corruption benchmarks for robustness evaluation, perceptual image quality assessment, and physically grounded analyses of imaging systems or real camera failures. Although these areas address closely related phenomena, they often use incompatible grouping schemes and backend specific severity definitions, making results difficult to compare across datasets, degradation sources, and tasks. We propose a causally grounded framework for organizing and interpreting image degradations across these settings. Instead of introducing new degradations or redefining existing benchmarks, we provide an interpretive representation and measurement layer that makes implicit assumptions explicit. Each degradation is described along two orthogonal axes: its dominant causal source in the imaging pipeline (environment, sensor/optics, ISP/renderer/codec, or transfer/system), and its resulting perceptual effect. This dual axis abstraction yields a compact taxonomy spanning algorithmic corruptions, perceptual distortions, and physically motivated imaging artifacts. To address inconsistent severity semantics without changing existing implementations, we introduce a lightweight severity measurement layer. For every degradation and each native severity level of a given backend, we quantify degradation strength using full reference image quality metrics: PSNR, SSIM, and LPIPS. This makes severity observable and comparable across sources while preserving native parameterizations. We demonstrate the framework through COCO Degradation, a taxonomy aligned benchmark for evaluating object detector robustness under diverse imaging conditions.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The paper proposes a causally grounded framework for organizing image degradations across communities studying synthetic corruptions, perceptual quality assessment, and physical imaging artifacts. Degradations are classified along two orthogonal axes: dominant causal source in the imaging pipeline (environment, sensor/optics, ISP/renderer/codec, or transfer/system) and resulting perceptual effect. A lightweight severity layer quantifies strength for each native severity level using PSNR, SSIM, and LPIPS to enable cross-source comparability without altering existing implementations. The framework is demonstrated via the COCO Degradation benchmark for object detector robustness evaluation.

Significance. If the taxonomy is shown to be exhaustive and free of overlaps while the severity metrics yield consistent, interpretable comparisons, the work could help unify robustness evaluations that currently use incompatible groupings and severity definitions. The decision to provide an interpretive layer rather than new degradations or benchmarks is a constructive strength that supports integration with existing datasets and tasks.

major comments (1)
  1. [Abstract] Abstract and framework description: The central claim that the four causal source categories form an exhaustive, mutually exclusive partition orthogonal to the perceptual-effect axis underpins the taxonomy's compactness and cross-source comparability. Degradations such as motion blur (potentially environment or sensor/optics) and JPEG artifacts (potentially ISP/renderer/codec or transfer/system) have ambiguous or multi-source origins. The manuscript must supply an explicit assignment procedure or boundary-case handling; absent this, the orthogonality premise risks overlaps that weaken the promised interpretive representation and severity observability.
minor comments (1)
  1. [Abstract] The abstract would benefit from a concise statement of how the COCO Degradation benchmark specifically differs from prior corruption suites in its use of the taxonomy.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback and positive assessment of the framework's potential to unify robustness evaluations. We address the major comment on causal source assignment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract and framework description: The central claim that the four causal source categories form an exhaustive, mutually exclusive partition orthogonal to the perceptual-effect axis underpins the taxonomy's compactness and cross-source comparability. Degradations such as motion blur (potentially environment or sensor/optics) and JPEG artifacts (potentially ISP/renderer/codec or transfer/system) have ambiguous or multi-source origins. The manuscript must supply an explicit assignment procedure or boundary-case handling; absent this, the orthogonality premise risks overlaps that weaken the promised interpretive representation and severity observability.

    Authors: We agree that an explicit assignment procedure strengthens the taxonomy. The manuscript already relies on the notion of 'dominant causal source' to resolve multi-source cases, but we will add a dedicated subsection in the revised Section 3 with a decision procedure and boundary-case examples. Motion blur will be assigned to environment when driven by scene or object motion and to sensor/optics for camera-induced effects such as shake or defocus. JPEG artifacts will be placed under ISP/renderer/codec because they arise during encoding and compression, while transfer/system covers post-encoding transmission losses. These rules, illustrated with concrete examples, preserve orthogonality to the perceptual-effect axis and make the framework easier to apply consistently. revision: yes

Circularity Check

0 steps flagged

No circularity: interpretive taxonomy and severity layer are self-contained

full rationale

The paper introduces a dual-axis framework (causal source categories and perceptual effects) plus a severity layer based on standard metrics (PSNR, SSIM, LPIPS) applied to existing backends. No equations, fitted parameters, or derivations are shown that reduce the taxonomy, orthogonality claim, or severity measures back to the paper's own inputs by construction. The four source categories are posited as an organizing premise rather than derived from prior self-citations or self-referential definitions. The work is presented as an independent interpretive layer over existing benchmarks, with no load-bearing self-citation chains or renaming of known results as new derivations.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on domain assumptions about the structure of the imaging pipeline and the applicability of full-reference metrics for severity without altering native implementations.

axioms (2)
  • domain assumption Image degradations have a dominant causal source that can be categorized into one of four stages: environment, sensor/optics, ISP/renderer/codec, or transfer/system.
    This defines the first axis of the proposed taxonomy as stated in the abstract.
  • domain assumption Full-reference image quality metrics (PSNR, SSIM, LPIPS) can quantify degradation strength for each native severity level while preserving the original backend parameterizations.
    This underpins the lightweight severity measurement layer introduced to address inconsistent severity semantics.

pith-pipeline@v0.9.0 · 5790 in / 1423 out tokens · 44313 ms · 2026-05-20T19:03:08.843448+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

65 extracted references · 65 canonical work pages

  1. [1]

    Benchmarking neural network robustness to common corruptions and perturbations,

    D. Hendrycks and T. Dietterich, “Benchmarking neural network robustness to common corruptions and perturbations,” inInternational Conference on Learning Representations, 2019. [Online]. Available: https://openreview.net/forum?id=HJz6tiCqYm

    work page 2019

  2. [2]

    Benchmarking robustness in object detection: Autonomous driving when winter is coming,

    C. Michaelis, B. Mitzkus, R. Geirhos, E. Rusak, O. Bringmann, A. S. Ecker, M. Bethge, and W. Brendel, “Benchmarking robustness in object detection: Autonomous driving when winter is coming,” inMachine Learning for Autonomous Driving (NeurIPS Workshop), 2019

    work page 2019

  3. [3]

    Arniqa: Learn- ing distortion manifold for image quality assessment,

    L. Agnolucci, L. Galteri, M. Bertini, and A. Del Bimbo, “Arniqa: Learn- ing distortion manifold for image quality assessment,” inProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2024, pp. 189–198

    work page 2024

  4. [4]

    Microsoft coco: Common objects in context,

    T.-Y . Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Doll ´ar, and C. L. Zitnick, “Microsoft coco: Common objects in context,” inEuropean conference on computer vision (ECCV). Springer, 2014, pp. 740–755

    work page 2014

  5. [5]

    Image quality assess- ment: from error visibility to structural similarity,

    Z. Wang, A. Bovik, H. Sheikh, and E. Simoncelli, “Image quality assess- ment: from error visibility to structural similarity,”IEEE Transactions on Image Processing, vol. 13, no. 4, pp. 600–612, 2004

    work page 2004

  6. [6]

    The unreasonable effectiveness of deep features as a perceptual metric,

    R. Zhang, P. Isola, A. A. Efros, E. Shechtman, and O. Wang, “The unreasonable effectiveness of deep features as a perceptual metric,” in CVPR, 2018

    work page 2018

  7. [7]

    Towards building self-aware object detectors via reliable uncertainty quantification and calibration,

    K. Oksuz, T. Joy, and P. K. Dokania, “Towards building self-aware object detectors via reliable uncertainty quantification and calibration,” inConference on Computer Vision and Pattern Recognition (CVPR), 2023

    work page 2023

  8. [8]

    Kadid-10k: A large-scale artificially distorted iqa database,

    H. Lin, V . Hosu, and D. Saupe, “Kadid-10k: A large-scale artificially distorted iqa database,” in2019 Eleventh International Conference on Quality of Multimedia Experience (QoMEX), 2019, pp. 1–3

    work page 2019

  9. [9]

    Benchmarking object detection robustness against real-world corruptions,

    J. Liu, Z. Wang, L. Ma, C. Fang, T. Bai, X. Zhang, J. Liu, and Z. Chen, “Benchmarking object detection robustness against real-world corruptions,”International Journal of Computer Vision, vol. 132, no. 10, pp. 4398–4416, Oct 2024. [Online]. Available: https://doi.org/10.1007/s11263-024-02096-6

    work page doi:10.1007/s11263-024-02096-6 2024

  10. [10]

    YOLOv10: Real-time end-to-end object detection,

    A. Wang, H. Chen, L. Liu, K. CHEN, Z. Lin, J. Han, and G. Ding, “YOLOv10: Real-time end-to-end object detection,” inThe Thirty- eighth Annual Conference on Neural Information Processing Systems,

    work page

  11. [11]

    Available: https://openreview.net/forum?id=tz83Nyb71l

    [Online]. Available: https://openreview.net/forum?id=tz83Nyb71l

    work page

  12. [12]

    Yolov9: Learning what you want to learn using programmable gradient information,

    C.-Y . Wang, I.-H. Yeh, and H.-Y . M. Liao, “Yolov9: Learning what you want to learn using programmable gradient information,” inEuropean Conference on Computer Vision (ECCV). Cham: Springer Nature Switzerland, 2024, pp. 1–21

    work page 2024

  13. [13]

    Ultralytics YOLO11,

    G. Jocher and J. Qiu, “Ultralytics YOLO11,” 2024. [Online]. Available: https://github.com/ultralytics/ultralytics

    work page 2024

  14. [14]

    DETRs Beat YOLOs on Real-time Object Detection,

    Y . Zhao, W. Lv, S. Xu, J. Wei, G. Wang, Q. Dang, Y . Liu, and J. Chen, “DETRs Beat YOLOs on Real-time Object Detection,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2024, pp. 16 965–16 974

    work page 2024

  15. [15]

    Quality-aware pre- trained models for blind image quality assessment,

    K. Zhao, K. Yuan, M. Sun, M. Li, and X. Wen, “Quality-aware pre- trained models for blind image quality assessment,” in2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 22 302–22 313

    work page 2023

  16. [16]

    Wise-od: Benchmarking robustness in infrared object detection,

    H. R. Medeiros, A. Belal, M. Aminbeidokhti, E. Granger, and M. Pedersoli, “Wise-od: Benchmarking robustness in infrared object detection,” 2025. [Online]. Available: https://arxiv.org/abs/2507.18925

    work page arXiv 2025

  17. [17]

    ImageNet: A Large-Scale Hierarchical Image Database,

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “ImageNet: A Large-Scale Hierarchical Image Database,” inCVPR09, 2009

    work page 2009

  18. [18]

    Learning multiple layers of features from tiny images,

    A. Krizhevsky, “Learning multiple layers of features from tiny images,” Tech. Rep., 2009

    work page 2009

  19. [19]

    The pascal visual object classes (voc) challenge,

    M. Everingham, L. Van Gool, C. K. Williams, J. Winn, and A. Zisser- man, “The pascal visual object classes (voc) challenge,”International journal of computer vision, vol. 88, pp. 303–338, 2010

    work page 2010

  20. [20]

    The cityscapes dataset for semantic urban scene understanding,

    M. Cordts, M. Omran, S. Ramos, T. Rehfeld, M. Enzweiler, R. Benen- son, U. Franke, S. Roth, and B. Schiele, “The cityscapes dataset for semantic urban scene understanding,” inProc. of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016

    work page 2016

  21. [21]

    Objects365: A large-scale, high-quality dataset for object detection,

    S. Shao, Z. Li, T. Zhang, C. Peng, G. Yu, J. Li, X. Zhang, and J. Sun, “Objects365: A large-scale, high-quality dataset for object detection,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019, pp. 8425–8434

    work page 2019

  22. [22]

    Bdd100k: A diverse driving dataset for heterogeneous multitask learning,

    F. Yu, H. Chen, X. Wang, W. Xian, Y . Chen, F. Liu, V . Madhavan, and T. Darrell, “Bdd100k: A diverse driving dataset for heterogeneous multitask learning,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2020

    work page 2020

  23. [23]

    Benchmarking robustness of 3d object detection to common corruptions in autonomous driving,

    Y . Dong, C. Kang, J. Zhang, Z. Zhu, Y . Wang, X. Yang, H. Su, X. Wei, and J. Zhu, “Benchmarking robustness of 3d object detection to common corruptions in autonomous driving,” in2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 1022– 1032

    work page 2023

  24. [24]

    Vision meets robotics: The kitti dataset,

    A. Geiger, P. Lenz, C. Stiller, and R. Urtasun, “Vision meets robotics: The kitti dataset,”International Journal of Robotics Research (IJRR), 2013

    work page 2013

  25. [25]

    nuscenes: A multimodal dataset for autonomous driving,

    H. Caesar, V . Bankiti, A. H. Lang, S. V ora, V . E. Liong, Q. Xu, A. Kr- ishnan, Y . Pan, G. Baldan, and O. Beijbom, “nuscenes: A multimodal dataset for autonomous driving,” inCVPR, 2020

    work page 2020

  26. [26]

    Scalability in perception for autonomous driving: Waymo open dataset,

    P. Sun, H. Kretzschmar, X. Dotiwalla, A. Chouard, V . Patnaik, P. Tsui, J. Guo, Y . Zhou, Y . Chai, B. Caine, V . Vasudevan, W. Han, J. Ngiam, H. Zhao, A. Timofeev, S. Ettinger, M. Krivokon, A. Gao, A. Joshi, Y . Zhang, J. Shlens, Z. Chen, and D. Anguelov, “Scalability in perception for autonomous driving: Waymo open dataset,” inProceedings of the IEEE/CV...

    work page 2020

  27. [27]

    3d common corruptions and data augmentation,

    O. F. Kar, T. Yeo, A. Atanov, and A. Zamir, “3d common corruptions and data augmentation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 18 963–18 974

    work page 2022

  28. [28]

    Multispectral fusion for object detection with cyclic fuse-and-refine blocks

    H. Zhang, E. Fromont, S. Lef `evre, and B. Avignon, “Multispectral fusion for object detection with cyclic fuse-and-refine blocks.” [Online]. Available: https://hal.archives-ouvertes.fr/hal-02872132

    work page

  29. [29]

    Free flir thermal dataset for algorithm training,

    FLIR (V2), “Free flir thermal dataset for algorithm training,” 2022,Available at https://www.flir.com/oem/adas/dataset/ european-regional-thermal-dataset/

    work page 2022

  30. [30]

    Llvip: A visible-infrared paired dataset for low-light vision,

    X. Jia, C. Zhu, M. Li, W. Tang, and W. Zhou, “Llvip: A visible-infrared paired dataset for low-light vision,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 3496–3504

    work page 2021

  31. [31]

    Benchmarking performance of object detection under image distortions in an uncontrolled environ- ment,

    A. Beghdadi, M. Mallem, and L. Beji, “Benchmarking performance of object detection under image distortions in an uncontrolled environ- ment,” in2022 IEEE International Conference on Image Processing (ICIP), 2022, pp. 2071–2075

    work page 2022

  32. [32]

    On interaction between augmentations and corruptions in natural corruption robustness,

    E. Mintun, A. Kirillov, and S. Xie, “On interaction between augmentations and corruptions in natural corruption robustness,” in Advances in Neural Information Processing Systems, M. Ranzato, A. Beygelzimer, Y . Dauphin, P. Liang, and J. W. Vaughan, Eds., vol. 34. Curran Associates, Inc., 2021, pp. 3571–3583. [Online]. Available: https://proceedings.neurip...

    work page 2021

  33. [33]

    CCC: Continuously changing corruptions,

    O. Press, S. Schneider, M. Kuemmerer, and M. Bethge, “CCC: Continuously changing corruptions,” inICML 2022 Shift Happens Workshop, 2022. [Online]. Available: https://openreview.net/forum?id= iK3ry72nCz8

    work page 2022

  34. [34]

    Coco- o: A benchmark for object detectors under natural distribution shifts,

    X. Mao, Y . Chen, Y . Zhu, D. Chen, H. Su, R. Zhang, and H. Xue, “Coco- o: A benchmark for object detectors under natural distribution shifts,” Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023

    work page 2023

  35. [35]

    Generalized out-of-distribution detection: A survey,

    J. Yang, K. Zhou, Y . Li, and Z. Liu, “Generalized out-of-distribution detection: A survey,”International Journal of Computer Vision, Jun

    work page

  36. [36]

    , author Zhou, K

    [Online]. Available: https://doi.org/10.1007/s11263-024-02117-4

    work page doi:10.1007/s11263-024-02117-4

  37. [37]

    The robustness of computer vision models against common corruptions: a survey,

    S. Wang, R. Veldhuis, and N. Strisciuglio, “The robustness of computer vision models against common corruptions: a survey,” 2023

    work page 2023

  38. [38]

    Multiscale structural similarity for image quality assessment,

    Z. Wang, E. Simoncelli, and A. Bovik, “Multiscale structural similarity for image quality assessment,” inThe Thrity-Seventh Asilomar Con- ference on Signals, Systems and Computers, 2003, vol. 2, 2003, pp. 1398–1402 V ol.2

    work page 2003

  39. [39]

    Fsim: A feature similarity index for image quality assessment,

    L. Zhang, L. Zhang, X. Mou, and D. Zhang, “Fsim: A feature similarity index for image quality assessment,”IEEE Transactions on Image Processing, vol. 20, no. 8, pp. 2378–2386, 2011

    work page 2011

  40. [40]

    No-reference image quality assessment in the spatial domain,

    A. Mittal, A. K. Moorthy, and A. C. Bovik, “No-reference image quality assessment in the spatial domain,”IEEE Transactions on Image Processing, vol. 21, pp. 4695–4708, 2012. [Online]. Available: https://api.semanticscholar.org/CorpusID:2927709

    work page 2012

  41. [41]

    Making a “completely blind

    A. Mittal, R. Soundararajan, and A. C. Bovik, “Making a “completely blind” image quality analyzer,”IEEE Signal Processing Letters, vol. 20, no. 3, pp. 209–212, 2013

    work page 2013

  42. [42]

    Blind image quality evaluation using perception based features,

    V . N, P. D, M. C. Bh, S. S. Channappayya, and S. S. Medasani, “Blind image quality evaluation using perception based features,” in 2015 Twenty First National Conference on Communications (NCC), 2015, pp. 1–6

    work page 2015

  43. [43]

    Blind image quality assessment: From natural scene statistics to perceptual quality,

    A. K. Moorthy and A. C. Bovik, “Blind image quality assessment: From natural scene statistics to perceptual quality,”IEEE Transactions on Image Processing, vol. 20, no. 12, pp. 3350–3364, 2011

    work page 2011

  44. [44]

    Musiq: Multi-scale image quality transformer,

    J. Ke, Q. Wang, Y . Wang, P. Milanfar, and F. Yang, “Musiq: Multi-scale image quality transformer,” in2021 IEEE/CVF International Conference on Computer Vision (ICCV), 2021, pp. 5128–5137

    work page 2021

  45. [45]

    Image quality assess- ment: Unifying structure and texture similarity,

    K. Ding, K. Ma, S. Wang, and E. P. Simoncelli, “Image quality assess- ment: Unifying structure and texture similarity,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 5, pp. 2567– 2581, 2022

    work page 2022

  46. [46]

    A survey on image quality assessment: Insights, analysis, and future outlook,

    C. Ma, Z. Shi, Z. Lu, S. Xie, F. Chao, and Y . Sui, “A survey on image quality assessment: Insights, analysis, and future outlook,” 2025. [Online]. Available: https://arxiv.org/abs/2502.08540

    work page arXiv 2025

  47. [47]

    Blind image quality assessment: A brief survey,

    M. Wang, “Blind image quality assessment: A brief survey,” 2023

    work page 2023

  48. [48]

    Image database tid2013: Peculiarities, results and perspectives,

    N. Ponomarenko, L. Jin, O. Ieremeiev, V . Lukin, K. Egiazarian, J. Astola, B. V ozel, K. Chehdi, M. Carli, F. Battisti, and C.-C. Jay Kuo, “Image database tid2013: Peculiarities, results and perspectives,” Signal Processing: Image Communication, vol. 30, pp. 57–77, 2015. [Online]. Available: https://www.sciencedirect.com/science/article/pii/ S0923596514001490

    work page 2015

  49. [49]

    Uhd- iqa benchmark database: Pushing the boundaries of blind photo quality assessment,

    V . Hosu, L. Agnolucci, O. Wiedemann, D. Iso, and D. Saupe, “Uhd- iqa benchmark database: Pushing the boundaries of blind photo quality assessment,” inComputer Vision – ECCV 2024 Workshops, A. Del Bue, C. Canton, J. Pont-Tuset, and T. Tommasi, Eds. Cham: Springer Nature Switzerland, 2025, pp. 467–482

    work page 2024

  50. [50]

    Going the extra mile in face image quality assessment: A novel database and model,

    S. Su, H. Lin, V . Hosu, O. Wiedemann, J. Sun, Y . Zhu, H. Liu, Y . Zhang, and D. Saupe, “Going the extra mile in face image quality assessment: A novel database and model,”IEEE Transactions on Multimedia, 2023

    work page 2023

  51. [51]

    Triangle orientation discrimination: the alternative to minimum resolvable temperature difference and minimum resolvable contrast,

    P. Bijl and J. M. Valeton, “Triangle orientation discrimination: the alternative to minimum resolvable temperature difference and minimum resolvable contrast,”Optical Engineering, vol. 37, no. 7, pp. 1976 – 1983, 1998. [Online]. Available: https://doi.org/10.1117/1.601904

    work page doi:10.1117/1.601904 1976

  52. [52]

    J. M. Lloyd,Thermal Imaging Systems. Plenum Press, 1975

    work page 1975

  53. [53]

    Impact of motion blur on recognition rates of CNN-based TOD classifier models,

    D. Wegner and S. Keßler, “Impact of motion blur on recognition rates of CNN-based TOD classifier models,” inElectro-Optical and Infrared Systems: Technology and Applications XX, D. L. Hickman, H. B ¨ursing, G. W. Kamerman, and O. Steinvall, Eds., vol. 12737, International Society for Optics and Photonics. SPIE, 2023, p. 127370L. [Online]. Available: https...

    work page doi:10.1117/12.2684366 2023

  54. [54]

    Thermal imager performance evaluation with recorded infrared images: Triangle Orientation Discrimination (TOD) classifier versus You Only Look Once (YOLO),

    D. Wegner, M. Wiehn, and S. Keßler, “Thermal imager performance evaluation with recorded infrared images: Triangle Orientation Discrimination (TOD) classifier versus You Only Look Once (YOLO),” inElectro-Optical and Infrared Systems: Technology and Applications XXII, D. L. Hickman, H. B ¨ursing, O. Steinvall, and P. J. Soan, Eds., vol. 13674, Internationa...

    work page doi:10.1117/12.3069905 2025

  55. [55]

    Semantic foggy scene understanding with synthetic data,

    C. Sakaridis, D. Dai, and L. Van Gool, “Semantic foggy scene understanding with synthetic data,”International Journal of Computer Vision, vol. 126, no. 9, pp. 973–992, Sep 2018. [Online]. Available: https://doi.org/10.1007/s11263-018-1072-8

    work page doi:10.1007/s11263-018-1072-8 2018

  56. [56]

    Hazydet: Open-source benchmark for drone-view object detec- tion with depth-cues in hazy scenes,

    C. Feng, Z. Chen, X. Li, C. Wang, J. Yang, M.-M. Cheng, Y . Dai, and Q. Fu, “Hazydet: Open-source benchmark for drone-view object detec- tion with depth-cues in hazy scenes,”arXiv preprint arXiv:2409.19833, 2025. APPENDIX SUPPLEMENTARYMATERIAL This supplementary material collects content that was omitted from the main paper for space reasons while remaini...

    work page arXiv 2025

  57. [57]

    (ImageNet-C/P)hendrycks/robustnessApache-2.0 License reported by repository meta- data

    work page

  58. [58]

    (Augmentation–Corruption)augmentation-corruptionMIT Repository states MIT license

    work page

  59. [59]

    (COCO-C variant)robust-detection-benchmarkMIT Repository states MIT license

    work page

  60. [60]

    (COCO-C variant)asharakeh/probdetApache-2.0 Repository states Apache-2.0 license

    work page

  61. [61]

    (Real-camera corruptions)real-world-benchmarkingn/adataset license CC BY 4.0; reposi- tory/license info not clearly stated on site

    work page

  62. [62]

    (3D Corruptions AD)3D Corruptions ADMIT Repository states MIT license

    work page

  63. [63]

    (3DCommonCorruptions)3DCommonCorruptionsAttribution- NonCommercial 4.0 International Repository states Attribution- NonCommercial 4.0 International license

    work page

  64. [64]

    (WiSE-OD)heitorrapela/wiseodn/ano license information

    work page

  65. [65]

    calib_images

    (EasyRobust / COCO-O)easyrobustApache-2.0 Repository states Apache-2.0 license. SUPPLEMENTARYMATERIAL: EXAMPLEUSAGE(I M G D E GLIBRARY): CALIBRATION ANDCANONICALDEGRADATION Theimdeglibrary separatesseverity calibrationfromdegradation application. Calibration maps backend-native severity levels to measured distortion strength, while degradation application...

    work page