pith. sign in

arxiv: 2605.19605 · v1 · pith:WAPBVC4Onew · submitted 2026-05-19 · 💻 cs.CV

deadtrees.earth-aerial: A Multi-Resolution Aerial Image Dataset for Tree Cover and Mortality Detection

Ayushi Sharma , Clemens Mosig , Lukas Drees , Salim Soltani , Janusch Vajna-Jehle , Aaron Sheppard , Belqis Ahmadi , Jonathan Schmid
show 4 more authors
Paul Neumeier Nathan Jacobs Jan Dirk Wegner Teja Kattenborn
This is my paper

Pith reviewed 2026-05-20 05:33 UTC · model grok-4.3

classification 💻 cs.CV
keywords aerial imagerytree mortalitytree cover segmentationmachine learning datasetsforest monitoringremote sensingcomputer visionbiome coverage
0
0 comments X p. Extension
pith:WAPBVC4O Add to your LaTeX paper What is a Pith Number? 
\usepackage{pith}
\pithnumber{WAPBVC4O}

Prints a linked pith:WAPBVC4O badge after your title and writes the identifier into PDF metadata. Compiles on arXiv with no extra files. Learn more

The pith

Two new open datasets enable joint segmentation of tree cover and mortality from global aerial imagery for the first time.

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

Forests worldwide face growing threats from climate change, fires, pests, and pathogens, making scalable monitoring of tree cover and mortality essential. Progress has been limited by the lack of harmonized, globally representative datasets suited for machine learning on centimeter-scale aerial imagery. This paper releases DTE-aerial-train, a training collection of 385K multi-resolution patches with audited pseudo-labels, and DTE-aerial-bench, a balanced test set of 25 orthoimages with expert annotations. The datasets together span tropical, temperate, boreal, and dryland biomes plus varied forest structures and mortality patterns. Baselines trained on the data raise mortality segmentation performance, including a notable lift in boreal regions.

Core claim

The paper introduces DTE-aerial-train, comprising 385K image patches of 1024x1024 pixels at 2.5-20 cm resolution with multi-class expert-annotated and audited pseudo-labels, and DTE-aerial-bench, a geographically balanced set of 25 orthoimages yielding 525 patches with high-quality expert labels for both tree cover and mortality. Both resources cover multiple biomes and support joint segmentation models that improve F1 scores for mortality detection across scales.

What carries the argument

The DTE-aerial-train and DTE-aerial-bench datasets, which supply multi-resolution aerial patches paired with labels for joint tree cover and mortality segmentation.

Load-bearing premise

The expert-annotated and audited pseudo-labels for tree cover and mortality are accurate and consistent enough across biomes, forest structures, and mortality patterns.

What would settle it

Independent expert re-annotation of a random subset of the benchmark patches that shows low agreement with the released labels would demonstrate that label quality is insufficient for reliable model training and evaluation.

Figures

Figures reproduced from arXiv: 2605.19605 by Aaron Sheppard, Ayushi Sharma, Belqis Ahmadi, Clemens Mosig, Jan Dirk Wegner, Janusch Vajna-Jehle, Jonathan Schmid, Lukas Drees, Nathan Jacobs, Paul Neumeier, Salim Soltani, Teja Kattenborn.

Figure 1
Figure 1. Figure 1: Dataset overview. (Top) Global, crowd-sourced aerial imagery from deadtrees.earth, representing one of the largest high-resolution collections for tree mortality. The training set spans tropical, temperate, boreal, and dryland biomes, capturing diverse canopy structures and forest conditions, while the benchmark is a curated subset emphasizing label quality. (Bottom) Distribution of sites and patches acros… view at source ↗
Figure 2
Figure 2. Figure 2: Qualitative examples of aerial imagery tiles and corresponding semantic masks (tree [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative comparison of segmentations on benchmark patches. Our method produces [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Multi-scale annotation and resampling in DTE-aerial-bench. Labels are created at 5 cm [PITH_FULL_IMAGE:figures/full_fig_p014_4.png] view at source ↗
read the original abstract

Forests worldwide are increasingly threatened by climate change and disturbances such as fire, pests, and pathogens, creating an urgent need for scalable monitoring of tree cover and tree mortality. Aerial imagery from drones and aircraft is a key data source for detailed and large-scale mapping of tree crowns and mortality. However, related progress is limited by the lack of globally representative, harmonized datasets for joint segmentation of tree cover and mortality. We introduce two novel, open, machine-learning-ready datasets to enable joint segmentation of tree cover and tree mortality from centimeter-scale aerial imagery for the first time at global scales. With DTE-aerial-train, we provide a training dataset comprising 385K image patches of size 1024x1024 pixels, with resolutions ranging from 2.5 to 20 cm. It includes multi-class expert-annotated and -audited pseudo-labels for tree cover and mortality. With DTE-aerial-bench, we provide a geographically balanced benchmark test set of 25 globally distributed orthoimages totaling 525 patches with high-quality expert annotations for both tree cover and mortality. Both the training and benchmark datasets span tropical, temperate, boreal, and dryland biomes and cover a wide range of forest structures and mortality patterns. Using the benchmark test set for evaluation, we establish strong reference baselines that improve mortality segmentation across all biomes and scales with significant gains in challenging regions, such as boreal forests, where the F1 score increases from 0.40 to 0.58 with around 45% relative improvement. All data, models, and code will be publicly released under permissive open-source licenses. An interactive visualization of the benchmark dataset is available at deadtrees.earth/releases/dte-aerial-bench.

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 manuscript introduces two novel open datasets for joint tree cover and mortality segmentation from centimeter-scale aerial imagery: DTE-aerial-train (385K 1024x1024 patches at 2.5-20 cm resolution with expert-annotated and audited pseudo-labels spanning tropical to boreal biomes) and DTE-aerial-bench (525 patches from 25 globally distributed orthoimages with high-quality expert annotations). It establishes baselines showing F1 gains for mortality detection (e.g., 0.40 to 0.58 in boreal forests) and commits to public release of data, models, and code.

Significance. If the labels hold, this release fills a clear gap in globally representative, harmonized training and benchmark data for high-resolution forest disturbance mapping, directly supporting scalable monitoring of climate-driven mortality across biomes. The combination of large training volume, multi-resolution coverage, and a geographically balanced expert-annotated test set, together with reproducible baselines, positions the work as a practical resource for the remote-sensing and computer-vision communities.

major comments (1)
  1. [Abstract] Abstract and dataset description: the central utility claim rests on DTE-aerial-train's 385K patches carrying sufficiently accurate multi-class labels for tree cover and mortality. The text describes these as 'expert-annotated and -audited pseudo-labels' but supplies no inter-annotator agreement statistics, audit sampling fractions, or per-biome/per-mortality-pattern error rates. Without such metrics, systematic label noise (e.g., under-detection of sparse mortality in dense tropical canopies or shadow-induced over-detection in boreal stands) cannot be quantified, directly affecting the reliability of any models trained on the data and the 'first at global scales' assertion.
minor comments (1)
  1. [Abstract] The manuscript could clarify the exact geographic balancing procedure used to select the 25 orthoimages for the benchmark set and whether any stratification by forest structure or mortality density was applied.

Simulated Author's Rebuttal

1 responses · 1 unresolved

We thank the referee for the constructive feedback emphasizing the need for greater transparency on label quality. We address the concern point by point below and commit to revisions that strengthen the dataset documentation without overstating current evidence.

read point-by-point responses

  1. Referee: [Abstract] Abstract and dataset description: the central utility claim rests on DTE-aerial-train's 385K patches carrying sufficiently accurate multi-class labels for tree cover and mortality. The text describes these as 'expert-annotated and -audited pseudo-labels' but supplies no inter-annotator agreement statistics, audit sampling fractions, or per-biome/per-mortality-pattern error rates. Without such metrics, systematic label noise (e.g., under-detection of sparse mortality in dense tropical canopies or shadow-induced over-detection in boreal stands) cannot be quantified, directly affecting the reliability of any models trained on the data and the 'first at global scales' assertion.

    Authors: We agree that the manuscript would benefit from additional detail on the labeling pipeline. The DTE-aerial-train labels combine initial expert annotations on a subset of images with model-generated pseudo-labels that were subsequently audited by the same expert team using standardized guidelines. Because the process relied on a single coordinated annotation team rather than multiple independent annotators, formal inter-annotator agreement statistics were not computed. In the revised manuscript we will add a new subsection under Dataset Construction that reports the audit sampling fraction (approximately 8 % of patches received full manual review) and provides a qualitative discussion of likely error modes by biome, drawing on the auditors' notes. We will also tone down the abstract phrasing to avoid implying that quantitative error rates have been measured. These changes directly respond to the referee's concern while preserving the claim that the dataset is the first harmonized multi-resolution, multi-biome resource of this scale. revision: yes

standing simulated objections not resolved
  • Quantitative per-biome and per-mortality-pattern error rates were not computed during the original auditing process and cannot be retroactively generated without re-annotating a statistically meaningful sample.

Circularity Check

0 steps flagged

No circularity; dataset release paper with independent benchmarks

full rationale

This manuscript is a data-release paper whose core contribution is the introduction of two new open datasets (DTE-aerial-train with 385K patches and DTE-aerial-bench with 25 orthoimages) plus empirical baselines evaluated on the held-out benchmark. No derivations, equations, fitted parameters, or predictions are claimed. The text describes expert-annotated pseudo-labels and reports F1 improvements on the new benchmark, but these are straightforward empirical evaluations against external test data rather than quantities that reduce to the paper's own inputs by construction. No self-citation chains, uniqueness theorems, or ansatzes are invoked to justify load-bearing steps. The paper is therefore self-contained against external benchmarks and receives the default non-circularity finding.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

As a dataset paper the contribution rests primarily on the quality and representativeness of the provided labels rather than on any modeling assumptions or free parameters.

axioms (1)
  • domain assumption Expert-annotated and audited pseudo-labels accurately represent tree cover and mortality across diverse biomes and forest structures
    The training and benchmark utility depends on this label quality; invoked in the abstract description of both datasets.

pith-pipeline@v0.9.0 · 5900 in / 1336 out tokens · 58534 ms · 2026-05-20T05:33:14.364548+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

41 extracted references · 41 canonical work pages · 2 internal anchors

  1. [1]

    In: 2019 IEEE 16th international symposium on biomedical imaging (ISBI 2019)

    Abraham, N., Khan, N.M.: A novel focal tversky loss function with improved attention u-net for lesion segmentation. In: 2019 IEEE 16th international symposium on biomedical imaging (ISBI 2019). pp. 683–687. IEEE (2019)

    work page 2019

  2. [2]

    Ecosphere6(8), 1–55 (2015)

    Allen, C.D., Breshears, D.D., McDowell, N.G.: On underestimation of global vulnerability to tree mortality and forest die-off from hotter drought in the anthropocene. Ecosphere6(8), 1–55 (2015)

    work page 2015

  3. [3]

    Environmental Data Science 3, e18 (2024)

    Allen, M.J., Moreno-Fernández, D., Ruiz-Benito, P., Grieve, S.W., Lines, E.R.: Low-cost tree crown dieback estimation using deep learning-based segmentation. Environmental Data Science 3, e18 (2024)

    work page 2024

  4. [4]

    arXiv preprint arXiv:2503.14273 (2025)

    Allen, M.J., Owen, H.J., Grieve, S.W., Lines, E.R.: Manual labelling artificially inflates deep learning-based segmentation performance on rgb images of closed canopy: validation using tls. arXiv preprint arXiv:2503.14273 (2025)

    work page arXiv 2025

  5. [5]

    Global Biogeochemical Cycles17(2) (2003)

    Bachelet, D., Neilson, R.P., Hickler, T., Drapek, R.J., Lenihan, J.M., Sykes, M.T., Smith, B., Sitch, S., Thonicke, K.: Simulating past and future dynamics of natural ecosystems in the united states. Global Biogeochemical Cycles17(2) (2003)

    work page 2003

  6. [6]

    Remote Sensing in Ecology and Conservation9(5), 641–655 (2023)

    Ball, J.G., Hickman, S.H., Jackson, T.D., Koay, X.J., Hirst, J., Jay, W., Archer, M., Aubry- Kientz, M., Vincent, G., Coomes, D.A.: Accurate delineation of individual tree crowns in tropical forests from aerial rgb imagery using mask r-cnn. Remote Sensing in Ecology and Conservation9(5), 641–655 (2023)

    work page 2023

  7. [7]

    In: Proceedings of the European conference on computer vision (ECCV)

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

    work page 2018

  8. [8]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Cheng, B., Misra, I., Schwing, A.G., Kirillov, A., Girdhar, R.: Masked-attention mask trans- former for universal image segmentation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 1290–1299 (2022)

    work page 2022

  9. [9]

    Nature communications15(1), 641 (2024)

    Cheng, Y ., Oehmcke, S., Brandt, M., Rosenthal, L., Das, A., Vrieling, A., Saatchi, S., Wagner, F., Mugabowindekwe, M., Verbruggen, W., et al.: Scattered tree death contributes to substantial forest loss in california. Nature communications15(1), 641 (2024)

    work page 2024

  10. [10]

    Remote Sensing of Environment311, 114283 (2024) 10

    Cloutier, M., Germain, M., Laliberté, E.: Influence of temperate forest autumn leaf phenology on segmentation of tree species from uav imagery using deep learning. Remote Sensing of Environment311, 114283 (2024) 10

    work page 2024

  11. [11]

    arXiv preprint arXiv:2602.02426 (2026)

    Duguay, S.O., Baudchon, H., Laliberté, E., Muller-Landau, H., Rivas-Torres, G., Ouaknine, A.: Selvamask: Segmenting trees in tropical forests and beyond. arXiv preprint arXiv:2602.02426 (2026)

    work page arXiv 2026

  12. [12]

    Earth System Science Data15(12), 5301–5369 (2023)

    Friedlingstein, P., O’sullivan, M., Jones, M.W., Andrew, R.M., Bakker, D.C., Hauck, J., Land- schützer, P., Le Quéré, C., Luijkx, I.T., Peters, G.P., et al.: Global carbon budget 2023. Earth System Science Data15(12), 5301–5369 (2023)

    work page 2023

  13. [13]

    Ecological Indicators145, 109627 (2022)

    Hagemann, N., Magdon, P., Schnell, S., Pommerening, A.: Analysing gap dynamics in forest canopies with landscape metrics based on multi-temporal airborne laser scanning surveys–a pilot study. Ecological Indicators145, 109627 (2022)

    work page 2022

  14. [14]

    Annual review of plant biology73(1), 673–702 (2022)

    Hartmann, H., Bastos, A., Das, A.J., Esquivel-Muelbert, A., Hammond, W.M., Martínez-Vilalta, J., McDowell, N.G., Powers, J.S., Pugh, T.A., Ruthrof, K.X., et al.: Climate change risks to global forest health: emergence of unexpected events of elevated tree mortality worldwide. Annual review of plant biology73(1), 673–702 (2022)

    work page 2022

  15. [15]

    Iakubovskii, P.: Segmentation models pytorch. github. GitHub repository https://github. com/qubvel/segmentation_models. pytorch (2019)

    work page 2019

  16. [16]

    ISPRS journal of photogrammetry and remote sensing173, 24–49 (2021)

    Kattenborn, T., Leitloff, J., Schiefer, F., Hinz, S.: Review on convolutional neural networks (cnn) in vegetation remote sensing. ISPRS journal of photogrammetry and remote sensing173, 24–49 (2021)

    work page 2021

  17. [17]

    Remote Sensing of Environment300, 113914 (2024)

    Khatri-Chhetri, P., van Wagtendonk, L., Hendryx, S.M., Kane, V .R.: Enhancing individual tree mortality mapping: The impact of models, data modalities, and classification taxonomy. Remote Sensing of Environment300, 113914 (2024)

    work page 2024

  18. [18]

    Quantifying the Carbon Emissions of Machine Learning

    Lacoste, A., Luccioni, A., Schmidt, V ., Dandres, T.: Quantifying the carbon emissions of machine learning. arXiv preprint arXiv:1910.09700 (2019)

    work page internal anchor Pith review Pith/arXiv arXiv 1910

  19. [19]

    Nature643(8074), 1203–1213 (2025)

    Migliavacca, M., Grassi, G., Bastos, A., Ceccherini, G., Ciais, P., Janssens-Maenhout, G., Lugato, E., Mahecha, M.D., Novick, K.A., Peñuelas, J., et al.: Securing the forest carbon sink for the european union’s climate ambition. Nature643(8074), 1203–1213 (2025)

    work page 2025

  20. [20]

    Authorea Preprints (2025)

    Möhring, J., Kattenborn, T., Mahecha, M.D., Cheng, Y ., Schwenke, M.B., Cloutier, M., Den- ter, M., Frey, J., Gassilloud, M., Göritz, A., et al.: Global, multi-scale standing deadwood segmentation in centimeter-scale aerial images. Authorea Preprints (2025)

    work page 2025

  21. [21]

    Mosig, C., Kattenborn, T., Loaiza, D.M., Vanja-Jehle, J., Brandt, J., Jacobs, N., Khanal, S., Xing, E., Schwartz, M., Muller-Landau, H.C., et al.: Sub-pixel mapping of disturbance and tree mortality dynamics from sentinel-2 time series around the globe (2026)

    work page 2026

  22. [22]

    Remote Sensing of Environment332, 115027 (2026)

    Mosig, C., Vajna-Jehle, J., Mahecha, M.D., Cheng, Y ., Hartmann, H., Montero, D., Junttila, S., Horion, S., Schwenke, M.B., Koontz, M.J., Maulud, K.N.A., Adu-Bredu, S., Al-Halbouni, D., Ali, M., Allen, M., Altman, J., Amorós, L., Angiolini, C., Astrup, R., Awada, H., Barrasso, C., Bartholomeus, H., Beck, P.S., Bozzini, A., Braun-Wimmer, J., Brede, B., Bre...

    work page doi:10.1016/j.rse.2025.115027 2026

  23. [23]

    New Phytologist245(6), 2377–2392 (2025)

    Network, I.T.M., Senf, C., Esquivel-Muelbert, A., Pugh, T.A., Anderegg, W.R., Anderson- Teixeira, K.J., Arellano, G., Beloiu Schwenke, M., Bentz, B.J., Boehmer, H.J., et al.: Towards a global understanding of tree mortality. New Phytologist245(6), 2377–2392 (2025)

    work page 2025

  24. [24]

    BioScience51(11), 933–938 (2001)

    Olson, D.M., Dinerstein, E., Wikramanayake, E.D., Burgess, N.D., Powell, G.V ., Underwood, E.C., D’amico, J.A., Itoua, I., Strand, H.E., Morrison, J.C., et al.: Terrestrial ecoregions of the world: A new map of life on earth: A new global map of terrestrial ecoregions provides an innovative tool for conserving biodiversity. BioScience51(11), 933–938 (2001)

    work page 2001

  25. [25]

    DINOv2: Learning Robust Visual Features without Supervision

    Oquab, M., Darcet, T., Moutakanni, T., V o, H., Szafraniec, M., Khalidov, V ., Fernandez, P., Haziza, D., Massa, F., El-Nouby, A., et al.: Dinov2: Learning robust visual features without supervision. arXiv preprint arXiv:2304.07193 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  26. [26]

    JHU press (2008)

    Perry, D.A., Oren, R., Hart, S.C.: Forest ecosystems. JHU press (2008)

    work page 2008

  27. [27]

    ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences2, 181–188 (2015)

    Polewski, P., Yao, W., Heurich, M., Krzystek, P., Stilla, U.: Detection of single standing dead trees from aerial color infrared imagery by segmentation with shape and intensity priors. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences2, 181–188 (2015)

    work page 2015

  28. [28]

    Bioscience58(6), 501–517 (2008)

    Raffa, K.F., Aukema, B.H., Bentz, B.J., Carroll, A.L., Hicke, J.A., Turner, M.G., Romme, W.H.: Cross-scale drivers of natural disturbances prone to anthropogenic amplification: the dynamics of bark beetle eruptions. Bioscience58(6), 501–517 (2008)

    work page 2008

  29. [29]

    In: International Conference on Medical image computing and computer-assisted intervention

    Ronneberger, O., Fischer, P., Brox, T.: U-net: Convolutional networks for biomedical image segmentation. In: International Conference on Medical image computing and computer-assisted intervention. pp. 234–241. Springer (2015)

    work page 2015

  30. [30]

    https://doi.org/10.35097/538, https://doi.org/ 10.35097/538

    Schiefer, F., Frey, J., Kattenborn, T.: Fortress: Forest tree species segmentation in very-high resolution uav-based orthomosaics (2022). https://doi.org/10.35097/538, https://doi.org/ 10.35097/538

    work page doi:10.35097/538 2022

  31. [31]

    ISPRS Journal of Photogrammetry and Remote Sensing170, 205–215 (2020)

    Schiefer, F., Kattenborn, T., Frick, A., Frey, J., Schall, P., Koch, B., Schmidtlein, S.: Mapping forest tree species in high resolution uav-based rgb-imagery by means of convolutional neural networks. ISPRS Journal of Photogrammetry and Remote Sensing170, 205–215 (2020)

    work page 2020

  32. [32]

    ISPRS Open Journal of Photogrammetry and Remote Sensing8, 100034 (2023)

    Schiefer, F., Schmidtlein, S., Frick, A., Frey, J., Klinke, R., Zielewska-Büttner, K., Junttila, S., Uhl, A., Kattenborn, T.: Uav-based reference data for the prediction of fractional cover of standing deadwood from sentinel time series. ISPRS Open Journal of Photogrammetry and Remote Sensing8, 100034 (2023)

    work page 2023

  33. [33]

    Forestry: An International Journal of Forest Research p

    Schiefer, F., Schmidtlein, S., Hartmann, H., Schnabel, F., Kattenborn, T.: Large-scale remote sensing reveals that tree mortality in germany appears to be greater than previously expected. Forestry: An International Journal of Forest Research p. cpae062 (2024)

    work page 2024

  34. [34]

    Proceedings of the national academy of sciences103(35), 13116–13120 (2006) 12

    Scholze, M., Knorr, W., Arnell, N.W., Prentice, I.C.: A climate-change risk analysis for world ecosystems. Proceedings of the national academy of sciences103(35), 13116–13120 (2006) 12

    work page 2006

  35. [35]

    Remote Sensing16(11), 1935 (2024)

    Troles, J., Schmid, U., Fan, W., Tian, J.: Bamforests: Bamberg benchmark forest dataset of individual tree crowns in very-high-resolution uav images. Remote Sensing16(11), 1935 (2024)

    work page 1935

  36. [36]

    Science349(6250), 814–818 (2015)

    Trumbore, S., Brando, P., Hartmann, H.: Forest health and global change. Science349(6250), 814–818 (2015)

    work page 2015

  37. [37]

    smithsonian tropical research institute (2023)

    Vasquez, V ., Cushman, K., Ramos, P., Williamson, C., Villareal, P., Gomez, C.L., Muller- Landau, H.: Barro colorado island 50-ha plot crown maps: manuallysegmented and instance segmented.(version 2). smithsonian tropical research institute (2023)

    work page 2023

  38. [38]

    Advances in neural information processing systems37, 49749–49767 (2024)

    Veitch-Michaelis, J., Cottam, A., Schweizer, D., Broadbent, E., Dao, D., Zhang, C., Almeyda Zambrano, A., Max, S.: Oam-tcd: A globally diverse dataset of high-resolution tree cover maps. Advances in neural information processing systems37, 49749–49767 (2024)

    work page 2024

  39. [39]

    In: The Thirty-ninth Annual Conference on Neural Information Processing Systems Datasets and Benchmarks Track

    Wang, Z., Li, C., Wang, R., Ma, L., Hurtt, G., Jia, X., Mai, G., Li, Z., Xie, Y .: Treefinder: A us-scale benchmark dataset for individual tree mortality monitoring using high-resolution aerial imagery. In: The Thirty-ninth Annual Conference on Neural Information Processing Systems Datasets and Benchmarks Track

    work page

  40. [40]

    Weinstein, B., Marconi, S., White, E.: Training data for the neontreeevaluation benchmark (2020)

    work page 2020

  41. [41]

    Xie, E., Wang, W., Yu, Z., Anandkumar, A., Alvarez, J.M., Luo, P.: Segformer: Simple and efficient design for semantic segmentation with transformers. Advances in neural information processing systems34, 12077–12090 (2021) A Data appendices A.1 Data preprocessing - DTE-aerial-train We applied a standardized preprocessing pipeline consisting of polygon lab...

    work page 2021