Pith. sign in

REVIEW 5 major objections 4 minor 47 references

GVCCS: A Dataset for Contrail Identification and Tracking on Visible Whole Sky Camera Sequences

T0 review · 5 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read GVCCS is an open, instance-level video dataset of contrails from a visible all-sky camera, with flight identifiers attached to each contrail and a unified Mask2Former baseline for segmentation and tracking.

desk verdict GVCCS is a genuinely new ground-based contrail video dataset with instance tracking and flight IDs, but the flight-attribution labels sit on an unvalidated fixed-altitude projection that deserves scrutiny. read the letter →

arxiv 2507.18330 v2 pith:YHDROIAJ submitted 2025-07-24 cs.CV cs.LG

Gabriel Jarry , Ramon Dalmau , Philippe Very , Franck Ballerini , Stefania-Denisa Bocu This is my paper
classification cs.CVcs.LG
keywords contrailsinstancesegmentationtemporaltrackingall-skycamerapanopticflightattributionvideodatasetclimateimpact
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper introduces GVCCS, an open dataset of 122 video sequences from a visible-range all-sky camera, totalling 24,228 frames, in which every contrail is individually annotated and tracked over time. The authors claim it is the first open-access, instance-level video dataset for contrail detection, segmentation, and tracking from ground-based sky imagery, and that it includes flight identifiers for contrails forming above the camera. This matters because physics-based contrail models need observational data with temporal continuity and flight attribution to be calibrated, and existing satellite datasets generally provide neither. The paper also supplies a Mask2Former-based panoptic segmentation baseline that performs semantic segmentation, instance segmentation, and temporal tracking in a single architecture.

What carries the argument

The load-bearing object is the GVCCS annotation structure: temporally consistent polygon masks with persistent instance identifiers, built by projecting all-sky images onto a 1024x1024 georeferenced grid and overlaying flight trajectories so annotators can mark 'new' contrails and assign flight IDs. The modelling machinery is Mask2Former, a transformer-based panoptic segmentation architecture whose masked attention focuses on regions near current predicted masks, which the authors extend to video clips and pair with a Hungarian-algorithm post-processing step that reconciles instance identifiers across overlapping clips. The combination allows a single model to output pixel-level class labels, per-instance masks, and consistent tracks, while the dataset's single- versus multi-polygon split isolates whether failures come from pixel detection or from grouping disconnected fragments.

What would settle it

Compare the projected aircraft trajectory with the observed contrail position for frames where the source flight's altitude is known from ADS-B data; if the horizontal offset between the projected track and the contrail grows as the flight's altitude differs from 10 km, the fixed-altitude projection is biasing the ground truth.

Watch

Extended reading notes

Core claim

The central claim is that a ground-based camera, combined with careful human annotation and flight-trajectory overlays, can produce a benchmark that tracks contrails as individual instances across their entire visible lifecycle. The dataset contains 4,651 labelled contrail instances drawn as 176,194 polygons, with 3,346 of those contrails linked to unique flight identifiers, and the authors demonstrate that a single Mask2Former model can segment contrail pixels, separate distinct contrails, and maintain instance identities over time. They further split the task into a multi-polygon formulation, where all fragments of one physical contrail are a single instance, and a single-polygon formulation, where each visible fragment is its own instance, allowing the benchmark to separate the visual grouping problem from flight-assisted association. The paper positions this resource as the missing human-labelled complement to synthetic and satellite-based datasets that lack per-frame tracking or flight attribution.

Load-bearing premise

The annotations and flight attribution rely on projecting the all-sky images onto a grid at a fixed cloud altitude of 10 km, so contrails forming much higher or lower could have their flight-track overlays misaligned with the visible trail.

Editorial extensions

If this is right

  • Researchers can train and evaluate models that follow an individual contrail from formation to dissipation, rather than only detecting contrail pixels in isolated frames.
  • The flight identifiers connect observed contrails to aircraft and engine parameters, enabling direct calibration of physical contrail models with real-world observations.
  • The single-polygon and multi-polygon formulations provide two well-defined benchmark tasks that separate pixel-level detection from the harder problem of grouping visually disconnected fragments.
  • The Mask2Former baseline, with the reported metrics, gives a reproducible reference point for comparing future contrail segmentation and tracking methods.
  • Because the dataset is released openly, other groups can add cameras in other regions and assemble a geographically diverse, temporally continuous record of contrail behaviour.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The authors leave implicit that the dataset's fixed-altitude projection could be corrected with per-flight altitude data, which would also make the attribution labels precise enough to train automated attribution models.
  • A testable extension is to measure contrail spreading and drift rates from the tracked masks and compare them against physics-based model outputs; the annotations already contain the necessary temporal resolution.
  • The multi-polygon instance task could serve as a general benchmark for grouping spatially disconnected object parts, since contrails fragment and fade in ways that resemble other thin, occluded structures.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 4 minor

Summary. The paper introduces GVCCS, a new open dataset of 122 visible-light ground-based all-sky video sequences (24,228 frames) with instance-level polygon annotations of contrails, temporal instance identities, and flight identifiers for contrails that form above the camera. It also presents baseline experiments using Mask2Former in image and video variants and a U-Net with a discriminative loss, evaluated with semantic segmentation metrics and COCO-style instance segmentation metrics adapted to an IoU range of [0.25, 0.75]. The central claims are that GVCCS is the first open-access, instance-level, temporally tracked, flight-attributed ground-based contrail video dataset and that the Mask2Former-based pipeline provides a unified segmentation and tracking baseline.

Significance. If the dataset construction is as reliable as described, GVCCS fills a clear gap in the literature: ground-based visible cameras provide high spatial and temporal resolution and capture contrails shortly after formation, and the instance-level temporal annotations with flight IDs support lifecycle studies and physical-model calibration. The authors provide a detailed multi-stage annotation protocol, a video-level train/test split, and an open release under CC BY 4.0 with COCO-format annotations and associated flight data. The baseline experiments are transparently described with public checkpoints and default hyperparameters. However, the significance is conditional on the validity of the labels, particularly the flight-attribution and new/old labels, which rest on a fixed-altitude geometric projection that is not validated, and on a quantitative demonstration of annotation reliability.

major comments (5)
  1. [Section 4.1] The georeferenced projection and the flight-trajectory overlays are computed at a fixed cloud altitude of 10 km, while contrails are stated to form at 8–12 km in Section 2.1. For a ray at zenith angle theta, the mismatch between an aircraft at altitude z and the 10 km projection plane is (10-z)*tan(theta); at z=8 km and theta=45 degrees this is 2 km, roughly 27 pixels at the stated 73 m/pixel scale, with a comparable error of opposite sign at z=12 km. Because the annotation tool overlays ADS-B trajectories to decide whether a contrail is 'new' and which flight produced it, this misalignment can bias or corrupt the flight IDs and the new/old distinction. No sensitivity analysis, no altitude distribution of attributed flights, and no independent validation of the attribution are reported. This is load-bearing for the dataset's flight-attribution claim; please add validation such as an altitude distribution, an error analysis, or a comparison with an independent attribution method.
  2. [Section 4.1] No inter-annotator agreement is reported. The multi-stage review protocol is described in detail, but the claim that the annotations are of high quality is not supported by any quantitative measure such as mask IoU between independent annotators, agreement on the new/old distinction, or agreement on flight ID assignment. Please report IAA statistics on a representative sample, or otherwise provide a quantitative reliability assessment.
  3. [Section 4.2 and Table 1] The dataset statistics are internally inconsistent. The prose states 176,194 annotated polygons and about 3,346 contrails associated with unique flight identifiers, while Table 1 reports 176,234 polygons and 3,354 unique flight IDs. Additionally, Section 4.1 states that each sequence contained between 60 and 480 images, whereas Table 1 reports a range of 41 to 600 images per sequence, and the introduction mentions durations of 20 minutes to 5 hours while Section 4.1 states 30 minutes to 4 hours. These numbers should be reconciled because the dataset description is a central part of the paper.
  4. [Section 6.2 and Appendix A] No tracking metrics are reported. The paper claims 'robust tracking of individual contrails over time' and presents a video-based model with a post-processing linking algorithm, but Tables 4 and 5 contain only per-frame segmentation metrics, and the text explicitly states that these scores do not account for flickering or instance identity consistency over time. Without measures such as IDF1, IDSW, MOTA, or track precision and recall, the tracking contribution is unverified. Please add tracking-specific evaluation metrics and also report the sensitivity of the results to the empirically chosen threshold tau=0.1 in Appendix A.
  5. [Section 6.2, Tables 4 and 5] The results are reported from a single 70/10/20 split with no error bars or confidence intervals, and the evaluation protocol uses an adapted IoU range of [0.25, 0.75] instead of the standard COCO range. The paper draws comparative conclusions such as 'the video-based model slightly outperforms' and 'Swin-L outperforms Swin-B' from differences that may be within noise. Please add uncertainty estimates, multiple seeds or split variations, and report at least one standard-COCO-range metric for comparability.
minor comments (4)
  1. [Throughout] There are several typos and grammatical issues, including 'doesn't explore all aspect' in the abstract, 'this paper present' in the introduction, 'we are deployed' in the conclusions, and 'HDSCAN' in Section 5.2, which should be 'HDBSCAN'.
  2. [Section 6.1] The description of the dataset split would benefit from reporting the number of videos and empty sequences in each of the training, validation, and test splits, since the balancing of empty sequences is claimed but not quantified.
  3. [Section 5.2] The sentence describing the clustering step, 'we use the HDSCAN algorithm, to find the clusters and a final k-means to associate outliers with closest cluster', is unclear and should be rewritten, and the algorithm name should be corrected.
  4. [Figure 4] The two-dimensional PCA embedding visualization would be more informative if the caption stated how many instances and pixels are shown and whether the model was trained on the displayed sample, since this affects the interpretability of the claimed cluster separation.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: the dataset, annotations, and baseline evaluations are self-contained; the minor self-citations are not load-bearing.

full rationale

The paper's main contribution is a new annotated dataset plus baseline segmentation models, not a derivation from first principles. There is no claimed derivation chain in which an output quantity is definitionally equal to an input quantity. The flight-attribution and 'new'/'old' labels rest on a geometric projection assumption (fixed 10 km cloud altitude, Section 4.1), but this is a correctness and validation concern, not a circularity: the labels are human annotations informed by flight trajectories, not predictions computed from the model under evaluation. The baseline models are trained and evaluated on a held-out video-level split (70-10-20, Section 6.1), which is standard supervised benchmark practice; no parameter is fitted to the test set and then reported as a predicted scientific result. The empirically selected tracking threshold (tau=0.1, Appendix A) and the adapted IoU range [0.25,0.75] (Section 6.2) affect reported scores and comparability, but they do not make the evaluation equivalent to the dataset construction or to the annotation inputs. The self-citations present—Jarry et al. (2025) for the dataset itself and Jarry et al. for the U-Net architecture and prior satellite segmentation work—are contextual or reference-architecture citations, and neither is used to justify a contested premise or to forbid alternative choices. Accordingly, no specific circular step can be exhibited, and the paper receives a low score reflecting only the presence of minor, non-load-bearing self-citations.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The dataset's core value rests on annotation quality, which depends on the fixed-altitude projection, flight-data accuracy, and human labeling skill. The model benchmarks add two empirically chosen parameters (IoU range and tracking threshold) but no new physical or algorithmic entities.

free parameters (2)
  • Tracking IoU threshold tau = 0.1 (recommended)
    Appendix A selects tau=0.1 empirically for matching instances across clips; the value affects track consistency and the resulting tracking behavior.
  • Evaluation IoU threshold range = 0.25 to 0.75
    Section 6.2 replaces the standard COCO range [0.50,0.95] with [0.25,0.75] to avoid penalizing thin contrails; this choice changes all reported AP/AR values and makes them not directly comparable to standard COCO scores.
assumptions (4)
  • domain assumption Fixed cloud altitude of 10 km for georeferenced projection
    Section 4.1 computes the 75 km x 75 km grid at a fixed cloud altitude. If contrails form at other altitudes, flight-trajectory overlays and pixel georeferencing may be misaligned, affecting annotation association.
  • domain assumption Accuracy and availability of flight trajectory (ADS-B) data above 15,000 ft
    Section 4.1 uses synchronized flight trajectories to label 'new' contrails and assign flight IDs. Incomplete or inaccurate ADS-B data would corrupt the flight-attribution ground truth.
  • domain assumption Human annotators can reliably distinguish contrails from natural clouds in visible fisheye images
    The annotation campaign relies on visual discrimination; Section 6.3 itself notes ambiguous cases and possible label noise, indicating this is a nontrivial assumption.
  • domain assumption Pretrained features from ImageNet, COCO, and YouTube-VIS transfer to contrail imagery
    Section 6.1 initializes all models from pretrained checkpoints, assuming transferred features help contrail segmentation; no ablation isolates the contribution of pretraining.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GVCCS: A Dataset for Contrail Identification and Tracking on Visible Whole Sky Camera Sequences." pith.science (2026). https://pith.science/paper/YHDROIAJ

@misc{pith2026250718330,
  author       = {Pith},
  title        = {Pith review of: GVCCS: A Dataset for Contrail Identification and Tracking on Visible Whole Sky Camera Sequences},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YHDROIAJ}},
  note         = {Machine review of arXiv:2507.18330}
}
read the original abstract

Aviation's climate impact includes not only CO2 emissions but also significant non-CO2 effects, especially from contrails. These ice clouds can alter Earth's radiative balance, potentially rivaling the warming effect of aviation CO2. Physics-based models provide useful estimates of contrail formation and climate impact, but their accuracy depends heavily on the quality of atmospheric input data and on assumptions used to represent complex processes like ice particle formation and humidity-driven persistence. Observational data from remote sensors, such as satellites and ground cameras, could be used to validate and calibrate these models. However, existing datasets don't explore all aspect of contrail dynamics and formation: they typically lack temporal tracking, and do not attribute contrails to their source flights. To address these limitations, we present the Ground Visible Camera Contrail Sequences (GVCCS), a new open data set of contrails recorded with a ground-based all-sky camera in the visible range. Each contrail is individually labeled and tracked over time, allowing a detailed analysis of its lifecycle. The dataset contains 122 video sequences (24,228 frames) and includes flight identifiers for contrails that form above the camera. As reference, we also propose a unified deep learning framework for contrail analysis using a panoptic segmentation model that performs semantic segmentation (contrail pixel identification), instance segmentation (individual contrail separation), and temporal tracking in a single architecture. By providing high-quality, temporally resolved annotations and a benchmark for model evaluation, our work supports improved contrail monitoring and will facilitate better calibration of physical models. This sets the groundwork for more accurate climate impact understanding and assessments.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

47 extracted references · 40 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  2. [2]

    , year 1953

    author Appleman, H. , year 1953 . title The formation of exhaust condensation trails by jet aircraft . journal Bulletin of the American Meteorological Society volume 34 , pages 14 -- 20 . :10.1175/1520-0477-34.1.14

    work page doi:10.1175/1520-0477-34.1.14 1953

  3. [3]

    , author Boucher, O

    author Borella, A. , author Boucher, O. , author Shine, K.P. , author Stettler, M. , author Tanaka, K. , author Teoh, R. , author Bellouin, N. , year 2024 . title The importance of an informed choice of co 2-equivalence metrics for contrail avoidance . journal Atmospheric Chemistry and Physics volume 24 , pages 9401--9417

    work page 2024

  4. [4]

    , author Neven, D

    author Brabandere, B.D. , author Neven, D. , author Gool, L.V. , year 2017 . title Semantic instance segmentation with a discriminative loss function . journal CoRR volume abs/1708.02551 . http://arxiv.org/abs/1708.02551, http://arxiv.org/abs/1708.02551 arXiv:1708.02551

    arXiv 2017

  5. [5]

    , author Massa, F

    author Carion, N. , author Massa, F. , author Synnaeve, G. , author Usunier, N. , author Kirillov, A. , author Zagoruyko, S. , year 2020 . title End-to-end object detection with transformers , in: booktitle European Conference on Computer Vision (ECCV) , pp. pages 213--229

    work page 2020

  6. [6]

    , author Papandreou, G

    author Chen, L.C. , author Papandreou, G. , author Schroff, F. , author Adam, H. , year 2017 . title Rethinking atrous convolution for semantic image segmentation . journal arXiv preprint arXiv:1706.05587

    arXiv 2017

  7. [7]

    , author Zhu, Y

    author Chen, L.C. , author Zhu, Y. , author Papandreou, G. , author Schroff, F. , author Adam, H. , year 2018 . title Encoder-decoder with atrous separable convolution for semantic image segmentation , in: booktitle Proceedings of the European conference on computer vision (ECCV) , pp. pages 801--818

    work page 2018

  8. [8]

    , author Choudhuri, A

    author Cheng, B. , author Choudhuri, A. , author Misra, I. , author Kirillov, A. , author Girdhar, R. , author Schwing, A.G. , year 2021 a. title Mask2former for video instance segmentation . journal arXiv preprint arXiv:2112.10764 note Extends Mask2Former to video by processing 3D segmentation volumes without modifying architecture or loss

    arXiv 2021

Show all 47 references
  1. [9]

    , author Misra, I

    author Cheng, B. , author Misra, I. , author Schwing, A.G. , author Kirillov, A. , author Girdhar, R. , year 2022 . title Masked-attention mask transformer for universal image segmentation , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Re...

    2022

  2. [10]

    , author Schwing, A.G

    author Cheng, B. , author Schwing, A.G. , author Kirillov, A. , year 2021 b. title Per-pixel classification is not all you need for semantic segmentation , in: booktitle Advances in Neural Information Processing Systems (NeurIPS)

    2021

  3. [11]

    , author Shapiro, M

    author Chevallier, R. , author Shapiro, M. , author Engberg, Z. , author Soler, M. , author Delahaye, D. , year 2023 . title Linear contrails detection, tracking and matching with aircraft using geostationary satellite and air traffic data . journal Aerospace volume 10 , pages 578

    2023

  4. [12]

    , author Liu, H

    author Dai, Z. , author Liu, H. , author Le, Q.V. , author Tan, M. , year 2021 . title Coatnet: Marrying convolution and attention for all data sizes . journal Advances in neural information processing systems volume 34 , pages 3965--3977

    2021

  5. [13]

    , author Minnis, P

    author Duda, D.P. , author Minnis, P. , author Khlopenkov, K. , author Chee, T.L. , author Boeke, R. , year 2013 . title Estimation of 2006 northern hemisphere contrail coverage using modis data . journal Geophysical Research Letters volume 40 , pages 612--617

    2013

  6. [14]

    , author Strauss, B

    author Forkert, T. , author Strauss, B. , author Wendling, P. , year 1993 . title A new algorithm for the automated detection of jet contrails from noaa-avhrr satellite images , in: booktitle Proc. of the 6th AVHRR Data Users' Meeting , organization EUMETSAT-Joint Research Cen...

    1993

  7. [15]

    , author Eastham, S.D

    author Fritz, T.M. , author Eastham, S.D. , author Speth, R.L. , author Barrett, S.R. , year 2020 . title The role of plume-scale processes in long-term impacts of aircraft emissions . journal Atmospheric Chemistry and Physics volume 20 , pages 5697--5727

    2020

  8. [16]

    , author Brand, E

    author Geraedts, S. , author Brand, E. , author Dean, T.R. , author Eastham, S. , author Elkin, C. , author Engberg, Z. , author Hager, U. , author Langmore, I. , author McCloskey, K. , author Ng, J.Y.H. , et al., year 2024 . title A scalable system to measure contrail formati...

    2024

  9. [17]

    , author Matthes, S

    author Gierens, K. , author Matthes, S. , author Rohs, S. , year 2020 . title How well can persistent contrails be predicted? journal Aerospace volume 7 . :10.3390/aerospace7120169

    2020 doi

  10. [18]

    , author Boucher, O

    author Gourgue, N. , author Boucher, O. , author Barth \`e s, L. , year 2025 . title A dataset of annotated ground-based images for the development of contrail detection algorithms . journal Data in Brief volume 59 , pages 111364

    2025

  11. [19]

    , author Gkioxari, G

    author He, K. , author Gkioxari, G. , author Doll \'a r, P. , author Girshick, R. , year 2017 . title Mask r-cnn , in: booktitle Proceedings of the IEEE International Conference on Computer Vision (ICCV) , pp. pages 2961--2969

    2017

  12. [20]

    , author Very, P

    author Jarry, G. , author Very, P. , author Ballerini, F. , author Dalmau, R. , year 2025 . title Gvccs : Ground visible camera contrail sequences . https://doi.org/10.5281/zenodo.16419651, :10.5281/zenodo.16419651

    2025 doi

  13. [21]

    , author Very, P

    author Jarry, G. , author Very, P. , author Heffar, A. , author Torjman-Levavasseur, V. , . title Deep semantic contrails segmentation of goes-16 satellite images: A hyperparameter exploration

  14. [22]

    , year 1955

    author Kuhn, H.W. , year 1955 . title The hungarian method for the assignment problem . journal Naval Research Logistics Quarterly volume 2 , pages 83--97

    1955

  15. [23]

    , year 2019

    author Kulik, L. , year 2019 . title Satellite-based detection of contrails using deep learning . Ph.D. thesis. Massachusetts Institute of Technology

    2019

  16. [24]

    , author Fahey, D.W

    author Lee, D.S. , author Fahey, D.W. , author Skowron, A. , author Allen, M.R. , author Burkhardt, U. , author Chen, Q. , author Doherty, S.J. , author Freeman, S. , author Forster, P.M. , author Fuglestvedt, J. , et al., year 2021 . title The contribution of global aviation ...

    2021

  17. [25]

    , author Maire, M

    author Lin, T. , author Maire, M. , author Belongie, S.J. , author Bourdev, L.D. , author Girshick, R.B. , author Hays, J. , author Perona, P. , author Ramanan, D. , author Doll \' a r, P. , author Zitnick, C.L. , year 2014 . title Microsoft COCO: common objects in context . j...

    2014 arXiv

  18. [26]

    , author Teoh, R

    author Low, J. , author Teoh, R. , author Ponsonby, J. , author Gryspeerdt, E. , author Shapiro, M. , author Stettler, M.E. , year 2025 . title Ground-based contrail observations: comparisons with reanalysis weather data and contrail model simulations . journal Atmospheric Mea...

    2025

  19. [27]

    , author Meyer, R

    author Mannstein, H. , author Meyer, R. , author Wendling, P. , year 1999 . title Operational detection of contrails from noaa-avhrr-data . journal International Journal of Remote Sensing volume 20 , pages 1641--1660

    1999

  20. [28]

    , author Geraedts, S.D

    author McCloskey, K.J.F. , author Geraedts, S.D. , author Jackman, B.H. , author Meijer, V.R. , author Brand, E.W. , author Fork, D.K. , author Platt, J.C. , author Elkin, C. , author Van Arsdale, C.H. , year 2021 . title A human-labeled landsat-8 contrails dataset , in: bookt...

    2021

  21. [29]

    , author Eastham, S.D

    author Meijer, V.R. , author Eastham, S.D. , author Waitz, I.A. , author Barrett, S.R. , year 2024 . title Contrail altitude estimation using goes-16 abi data and deep learning . journal Atmospheric Measurement Techniques volume 17 , pages 6145--6162

    2024

  22. [30]

    , author Kulik, L

    author Meijer, V.R. , author Kulik, L. , author Eastham, S.D. , author Allroggen, F. , author Speth, R.L. , author Karaman, S. , author Barrett, S.R. , year 2022 . title Contrail coverage over the united states before and during the covid-19 pandemic . journal Environmental Re...

    2022

  23. [31]

    , author McCloskey, K

    author Ng, J.Y.H. , author McCloskey, K. , author Cui, J. , author Meijer, V.R. , author Brand, E. , author Sarna, A. , author Goyal, N. , author Van Arsdale, C. , author Geraedts, S. , year 2023 . title Opencontrails: Benchmarking contrail detection on goes-16 abi . journal a...

    2023 arXiv

  24. [32]

    , author Jiang, L

    author Northcutt, C. , author Jiang, L. , author Chuang, I. , year 2021 . title Confident learning: Estimating uncertainty in dataset labels . journal J. Artif. Int. Res. volume 70 , pages 1373–1411 . :10.1613/jair.1.12125

    2021 doi

  25. [33]

    , author Garc \' a-Heras, J

    author Ortiz, I. , author Garc \' a-Heras, J. , author Jafarimoghaddam, A. , author Soler, M. , year 2025 . title Enhancing goes-16 contrail segmentation through ensemble neural network modeling and optical flow corrections . journal Authorea Preprints

    2025

  26. [34]

    , year 2007

    author Pratt, W.K. , year 2007 . title Digital image processing: PIKS Scientific inside . volume volume 4 . publisher Wiley Online Library

    2007

  27. [35]

    , author Baruch, E.B

    author Ridnik, T. , author Baruch, E.B. , author Noy, A. , author Zelnik - Manor, L. , year 2021 . title Imagenet-21k pretraining for the masses . journal CoRR volume abs/2104.10972 . http://arxiv.org/abs/2104.10972 arXiv:2104.10972

    2021 arXiv

  28. [36]

    , author Dubot, T

    author Riggi-Carrolo, E. , author Dubot, T. , author Sarrat, C. , author Bedouet, J. , year 2023 . title Ai-driven identification of contrail sources: Integrating satellite observation and air traffic data . journal Journal of Open Aviation Science volume 1

    2023

  29. [37]

    , author Fischer, P

    author Ronneberger, O. , author Fischer, P. , author Brox, T. , year 2015 . title U-net: Convolutional networks for biomedical image segmentation , in: booktitle International Conference on Medical image computing and computer-assisted intervention , organization Springer . pp...

    2015

  30. [38]

    , author Meijer, V

    author Sarna, A. , author Meijer, V. , author Chevallier, R. , author Duncan, A. , author McConnaughay, K. , author Geraedts, S. , author McCloskey, K. , year 2025 . title Benchmarking and improving algorithms for attributing satellite-observed contrails to flights . journal E...

    2025

  31. [39]

    , year 2012

    author Schumann, U. , year 2012 . title A contrail cirrus prediction model . journal Geoscientific Model Development volume 5 , pages 543--580 . https://gmd.copernicus.org/articles/5/543/2012/, :10.5194/gmd-5-543-2012

    2012 doi

  32. [40]

    , author Hempel, R

    author Schumann, U. , author Hempel, R. , author Flentje, H. , author Garhammer, M. , author Graf, K. , author Kox, S. , author L \"o sslein, H. , author Mayer, B. , year 2013 . title Contrail study with ground-based cameras . journal Atmospheric Measurement Techniques volume ...

    2013

  33. [41]

    , author Roosenbrand, E

    author Sun, J. , author Roosenbrand, E. , year 2025 . title Few-shot contrail segmentation in remote sensing imagery with loss function in hough space . journal IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing

    2025

  34. [42]

    , author Engberg, Z

    author Teoh, R. , author Engberg, Z. , author Schumann, U. , author Voigt, C. , author Shapiro, M. , author Rohs, S. , author Stettler, M. , year 2023 . title Global aviation contrail climate effects from 2019 to 2021 . journal EGUsphere volume 2023 , pages 1--32

    2023

  35. [43]

    , author Schumann, U

    author Teoh, R. , author Schumann, U. , author Stettler, M.E. , year 2020 . title Beyond contrail avoidance: Efficacy of flight altitude changes to minimise contrail climate forcing . journal Aerospace volume 7 , pages 121

    2020

  36. [44]

    , author Talebi, H

    author Tu, Z. , author Talebi, H. , author Zhang, H. , author Yang, F. , author Milanfar, P. , author Bovik, A. , author Li, Y. , year 2022 . title Maxvit: Multi-axis vision transformer , in: booktitle European conference on computer vision , organization Springer . pp. pages 459--479

    2022

  37. [45]

    , author Ehrmanntraut, R

    author Van Huffel, J. , author Ehrmanntraut, R. , author Croes, D. , year 2025 . title Contrail detection and classification using computer vision with ground-based cameras , in: booktitle 2025 Integrated Communications, Navigation and Surveillance Conference (ICNS) , organiza...

    2025

  38. [46]

    , author Mannstein, H

    author Vazquez-Navarro, M. , author Mannstein, H. , author Mayer, B. , year 2010 . title An automatic contrail tracking algorithm . journal Atmospheric Measurement Techniques volume 3 , pages 1089--1101

    2010

  39. [47]

    , author Fan, Y

    author Yang, L. , author Fan, Y. , author Xu, N. , year 2019 . title Video instance segmentation , in: booktitle ICCV

    2019

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.