pith. machine review for the scientific record. sign in

arxiv: 2604.20000 · v1 · submitted 2026-04-21 · 💻 cs.CV

Recognition: unknown

RareSpot+: A Benchmark, Model, and Active Learning Framework for Small and Rare Wildlife in Aerial Imagery

Bowen Zhang , Jesse T. Boulerice , Charvi Mendiratta , Nikhil Kuniyil , Satish Kumar , Hila Shamon , B. S. Manjunath
Authors on Pith no claims yet

Pith reviewed 2026-05-10 02:17 UTC · model grok-4.3

classification 💻 cs.CV
keywords aerial imageryobject detectionwildlife monitoringactive learningprairie dogsrare speciesmulti-scale consistencygeospatial priors
0
0 comments X

The pith

RareSpot+ improves detection of small rare wildlife like prairie dogs in aerial imagery by 35 percent while using just 1.7 percent of tiles for new labels.

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

The paper sets out to show that adding a multi-scale consistency loss to align features across detection heads, plus context-aware augmentation and geospatial active learning that exploits animal-burrow location patterns, can overcome the core difficulties of spotting tiny, sparse animals against complex backgrounds in large aerial photos. Traditional detectors miss these objects because they occupy only about 30 pixels and blend with surroundings, while the cost of expert labeling across square kilometers quickly becomes prohibitive. If the approach works, conservation teams could survey ecologically important species over far larger areas at lower cost and directly turn the detections into quantitative maps of clustering and species co-occurrence.

Core claim

RareSpot+ combines a multi-scale consistency loss that aligns intermediate feature maps from multiple detection heads to sharpen localization of small objects, context-aware augmentation that generates ecologically plausible hard examples, and a geospatial active learning module that uses spatial priors between prairie dogs and burrows together with test-time augmentation and a meta-uncertainty model to pick the most informative tiles. On a 2 km² aerial dataset this yields a 35.2 percent relative gain in mAP@50 (absolute +0.13) over baseline detectors. The active learning component adds a further 14.5 percent to prairie dog average precision when only 1.7 percent of the unlabeled tiles are标注

What carries the argument

The multi-scale consistency loss that forces alignment of feature maps across detection heads for small-object localization, paired with geospatially guided active learning that selects tiles using animal-burrow spatial priors and uncertainty estimates.

Load-bearing premise

The multi-scale consistency loss and geospatial priors deliver the reported gains without heavy per-dataset tuning and that the transferability seen on the tested wildlife sets extends to new species and imaging conditions.

What would settle it

Apply RareSpot+ unchanged to a new 2 km² aerial dataset of a different small rare species and check whether mAP@50 gain drops below 15 percent or whether active learning needs more than 5 percent of tiles to reach a 10 percent AP lift.

read the original abstract

Automated wildlife monitoring from aerial imagery is vital for conservation but remains limited by two persistent challenges: the difficulty of detecting small, rare species and the high cost of large-scale expert annotation. Prairie dogs exemplify this problem -- they are ecologically important yet appear tiny, sparsely distributed, and visually indistinct from their surroundings, posing a severe challenge for conventional detection models. To overcome these limitations, we present RareSpot+, a detection framework that integrates multi-scale consistency learning, context-aware augmentation, and geospatially guided active learning to address these issues. A novel multi-scale consistency loss aligns intermediate feature maps across detection heads, enhancing localization of small (approx. 30 pixels wide) objects without architectural changes, while context-aware augmentation improves robustness by synthesizing hard, ecologically plausible examples. A geospatial active learning module exploits domain-specific spatial priors linking prairie dogs and burrows, together with test-time augmentation and a meta-uncertainty model, to reduce redundant labeling. On a 2 km^2 aerial dataset, RareSpot+ improves detection over the baseline mAP@50 by +35.2% (absolute +0.13). Cross-dataset tests on HerdNet, AED, and several other wildlife benchmarks demonstrate robust detector-level transferability. The active learning module further boosts prairie dog AP by 14.5% using an annotation budget of just 1.7% of the unlabeled tiles. Beyond detection, RareSpot+ enables spatial ecological analyses such as clustering and co-occurrence, linking vision-based detection with quantitative ecology.

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

3 major / 3 minor

Summary. The paper introduces RareSpot+, a detection framework for small and rare wildlife (exemplified by prairie dogs) in aerial imagery. It combines a multi-scale consistency loss that aligns intermediate feature maps across detection heads to improve localization of ~30-pixel objects, context-aware augmentation to synthesize hard examples, and a geospatial active learning module that incorporates burrow spatial priors, test-time augmentation, and meta-uncertainty to minimize annotation effort. On a 2 km² prairie-dog aerial dataset the method reports an absolute +0.13 mAP@50 gain (+35.2% relative) over baseline; cross-dataset detector transfer is shown on HerdNet, AED and other wildlife benchmarks; the active-learning component yields a further +14.5% AP lift at 1.7% annotation budget. The work also connects detections to downstream ecological analyses such as clustering and co-occurrence.

Significance. If the reported gains prove robustly attributable to the proposed components rather than tuning artifacts, the work would meaningfully advance automated conservation monitoring by tackling the twin difficulties of small-object detection and expensive expert labeling. The domain-specific geospatial priors and their integration with active learning are a concrete strength, and the explicit linkage to quantitative ecology broadens impact beyond pure computer vision. Cross-dataset detector results provide some evidence of transferability, though the active-learning module itself is not similarly tested.

major comments (3)
  1. [Experiments / prairie-dog dataset evaluation] The central empirical claim—an absolute +0.13 mAP@50 improvement on the 2 km² prairie-dog dataset—is load-bearing for the paper’s contribution. No ablation experiments are described that isolate the multi-scale consistency loss or context-aware augmentation (e.g., by removing each term and re-training the baseline detector under identical hyper-parameter search). Without these controls it remains possible that comparable gains arise from standard tuning rather than the novel losses.
  2. [Active Learning Module Evaluation] The active-learning result (+14.5% AP at 1.7% annotation budget) attributes efficiency to the combination of burrow spatial priors and meta-uncertainty. The manuscript provides no comparison against standard uncertainty sampling or random selection baselines that lack the geospatial prior; therefore the contribution of the domain-specific module cannot be quantified, and generalizability to other rare-species settings without analogous priors is untested.
  3. [Cross-Dataset Experiments] Cross-dataset transferability is asserted for the detector on HerdNet, AED and additional wildlife benchmarks, yet the active-learning framework is evaluated exclusively on the prairie-dog data. This asymmetry weakens the claim that the full RareSpot+ pipeline (including the geospatial module) transfers robustly.
minor comments (3)
  1. [Results] Numeric results are presented without error bars, standard deviations across random seeds, or statistical significance tests; adding these would allow readers to assess stability of the +0.13 mAP and +14.5% AP figures.
  2. [Method / Multi-scale Consistency Loss] The multi-scale consistency loss is described at a high level; an explicit equation (e.g., the precise form of the alignment term between feature maps) would improve reproducibility.
  3. [Abstract / Experiments] The abstract refers to “several other wildlife benchmarks” beyond HerdNet and AED; an explicit list or table in the experiments section would clarify the scope of the transferability claims.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. We address each major comment below, indicating the revisions we will incorporate to strengthen the empirical validation.

read point-by-point responses

  1. Referee: [Experiments / prairie-dog dataset evaluation] The central empirical claim—an absolute +0.13 mAP@50 improvement on the 2 km² prairie-dog dataset—is load-bearing for the paper’s contribution. No ablation experiments are described that isolate the multi-scale consistency loss or context-aware augmentation (e.g., by removing each term and re-training the baseline detector under identical hyper-parameter search). Without these controls it remains possible that comparable gains arise from standard tuning rather than the novel losses.

    Authors: We agree that component-wise ablations under matched hyperparameter search are required to rigorously attribute gains to the proposed losses rather than tuning. The manuscript currently presents the overall improvement and cross-dataset detector transfer as supporting evidence, but lacks these controls. In the revision we will add explicit ablation studies on the prairie-dog dataset that remove the multi-scale consistency loss and the context-aware augmentation individually while re-training the baseline under the same search protocol. revision: yes

  2. Referee: [Active Learning Module Evaluation] The active-learning result (+14.5% AP at 1.7% annotation budget) attributes efficiency to the combination of burrow spatial priors and meta-uncertainty. The manuscript provides no comparison against standard uncertainty sampling or random selection baselines that lack the geospatial prior; therefore the contribution of the domain-specific module cannot be quantified, and generalizability to other rare-species settings without analogous priors is untested.

    Authors: We concur that baselines without the geospatial prior are needed to quantify its incremental value. We will add comparisons against standard uncertainty sampling and random selection on the prairie-dog dataset. Regarding generalizability, the module is designed around domain-specific priors; we will expand the discussion to clarify how the framework can be instantiated in other rare-species settings where analogous spatial or ecological priors can be obtained or learned. revision: yes

  3. Referee: [Cross-Dataset Experiments] Cross-dataset transferability is asserted for the detector on HerdNet, AED and additional wildlife benchmarks, yet the active-learning framework is evaluated exclusively on the prairie-dog data. This asymmetry weakens the claim that the full RareSpot+ pipeline (including the geospatial module) transfers robustly.

    Authors: The active-learning module relies on prairie-dog-specific burrow spatial priors that are unavailable in HerdNet, AED, or the other benchmarks. Consequently the full pipeline (detector + AL) is evaluated only where those priors exist, while the detector components alone are tested for cross-dataset transfer. We will revise the text to explicitly distinguish these scopes and avoid implying that the complete pipeline transfers without equivalent priors. revision: partial

Circularity Check

0 steps flagged

No circularity: empirical claims rest on reported experiments without self-referential derivations.

full rationale

The paper introduces a detection framework with multi-scale consistency loss, context-aware augmentation, and geospatial active learning, then reports mAP@50 gains (+0.13 absolute) and active-learning efficiency on a 2 km² prairie-dog dataset plus cross-dataset transfer. No equations, derivations, or parameter-fitting steps appear in the abstract or described content. Performance numbers are presented as direct experimental outcomes rather than quantities that reduce to fitted inputs or self-citations by construction. No uniqueness theorems, ansatzes smuggled via prior work, or renamed known results are invoked. The derivation chain is therefore self-contained and non-circular.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no identifiable free parameters, axioms, or invented entities. No equations or implementation details are supplied that would reveal fitted scales, ad-hoc assumptions, or new postulated constructs.

pith-pipeline@v0.9.0 · 5605 in / 1261 out tokens · 42864 ms · 2026-05-10T02:17:54.372245+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

75 extracted references · 28 canonical work pages · 3 internal anchors

  1. [1]

    , Zhang, C

    ash2019deep APACrefauthors Ash, J T. , Zhang, C. , Krishnamurthy, A. , Langford, J. \ Agarwal, A. APACrefauthors \ 2020 . Deep Batch Active Learning by Diverse, Uncertain Gradient Lower Bounds. Deep batch active learning by diverse, uncertain gradient lower bounds

    2020

  2. [2]

    Bisque: Bioimage software and services platform

    bisque Bisque: Bioimage Software and Services Platform. Bisque: Bioimage software and services platform. . https://bisque2.ece.ucsb.edu/client_service/ . Accessed: 2025

    2025

  3. [3]

    YOLOv4: Optimal Speed and Accuracy of Object Detection

    bochkovskiy2020yolov4 APACrefauthors Bochkovskiy, A. , Wang, C Y. \ Liao, H Y M. APACrefauthors \ 2020 . YOLOv4: Optimal Speed and Accuracy of Object Detection Yolov4: Optimal speed and accuracy of object detection . arXiv preprint arXiv:2004.10934

    work page internal anchor Pith review arXiv 2020

  4. [4]

    APACrefauthors \ 2000

    opencv_library APACrefauthors Bradski, G. APACrefauthors \ 2000 . The OpenCV Library The OpenCV Library . Dr. Dobb's Journal of Software Tools

    2000

  5. [5]

    , K\" o nig, C

    Brust2019ALOD APACrefauthors Brust, C A. , K\" o nig, C. , Denzler, M. , Denzler, J. \ Rodner, E. APACrefauthors \ 2019 . Active Learning for Deep Object Detection. Active learning for deep object detection

    2019

  6. [6]

    , Anderson, D R

    Buckland2001Book APACrefauthors Buckland, S T. , Anderson, D R. , Burnham, K P. , Laake, J L. , Borchers, D L. \ Thomas, L. APACrefauthors \ 2001 . Introduction to Distance Sampling: Estimating Abundance of Biological Populations Introduction to distance sampling: Estimating abundance of biological populations . Oxford, UK Oxford University Press

    2001

  7. [7]

    Xinlei Chen, Hao Fang, Tsung-yi Lin, Ramakrishna Vedantam, Saurabh Gupta, Piotr Dollar, and C Lawrence Zitnick

    carion2020detr APACrefauthors Carion, N. , Massa, F. , Synnaeve, G. , Usunier, N. , Kirillov, A. \ Zagoruyko, S. APACrefauthors \ 2020 1 . End-to-End Object Detection with Transformers. End-to-end object detection with transformers. APACrefURL https://arxiv.org/abs/2005.12872 APACrefURL

    work page arXiv 2020

  8. [8]

    , Massa, F

    carion2020end APACrefauthors Carion, N. , Massa, F. , Synnaeve, G. , Usunier, N. , Kirillov, A. \ Zagoruyko, S. APACrefauthors \ 2020 2 . End-to-end object detection with transformers. End-to-end object detection with transformers. Springer

    2020

  9. [9]

    APACrefauthors \ 1974

    Caughley1974JWM APACrefauthors Caughley, G. APACrefauthors \ 1974 . Bias in Aerial Survey Bias in aerial survey . The Journal of Wildlife Management 38 4 921--933

    1974

  10. [10]

    \ Francis, C M

    chabot2018canada APACrefauthors Chabot, D. \ Francis, C M. APACrefauthors \ 2018 . Canada Goose Nesting Ecology and Management: Combining Drone Imagery and Ground Surveys to Improve Population Monitoring Canada goose nesting ecology and management: Combining drone imagery and ground surveys to improve population monitoring . Remote Sensing in Ecology and ...

    work page doi:10.1002/rse2.59 2018

  11. [11]

    \ Lightfoot, D C

    DavidsonLightfoot2006Ecography APACrefauthors Davidson, A D. \ Lightfoot, D C. APACrefauthors \ 2006 . Keystone Rodent Interactions: Prairie Dogs and Kangaroo Rats Structure the Biotic Composition of a Desertified Grassland Keystone rodent interactions: Prairie dogs and kangaroo rats structure the biotic composition of a desertified grassland . Ecography ...

    work page doi:10.1111/j.2006.0906-7590.04699.x 2006

  12. [12]

    , Bly, K L S

    Delparte APACrefauthors Delparte, D M. , Bly, K L S. , Stone, T. , Olimb, S K. , Kinsey, M J. , Belt, M. \ Calton, T F. APACrefauthors \ 2019 . sUAS for Wildlife Conservation – Assessing Habitat Quality of the Endangered Black-Footed Ferret suas for wildlife conservation – assessing habitat quality of the endangered black-footed ferret . Applications of S...

    2019

  13. [13]

    , Foucher, S

    Delplanque2023ISPRS APACrefauthors Delplanque, A. , Foucher, S. , Th \'e au, J. , Bussi \`e re, E. , Vermeulen, C. \ Lejeune, P. APACrefauthors \ 2023 . From crowd to herd counting: How to precisely detect and count African mammals using aerial imagery and deep learning? From crowd to herd counting: How to precisely detect and count african mammals using ...

    work page doi:10.1016/j.isprsjprs.2023.01.025 2023

  14. [14]

    Journal of Wildlife Management 69, 1691 (2005) https://doi.org/10.2193/0022-541X(2005)69[1691:HSBEBA]2.0.CO;2

    Dinsmore2005JWM APACrefauthors Dinsmore, S J. , White, G C. \ Knopf, F L. APACrefauthors \ 2005 . Mountain Plover Population Responses to Black-Tailed Prairie Dogs in Montana Mountain plover population responses to black-tailed prairie dogs in montana . The Journal of Wildlife Management 69 4 1546--1553 . APACrefDOI doi:10.2193/0022-541X(2005)69[1546:MPPR...

    work page doi:10.2193/0022-541x(2005)69 2005

  15. [15]

    , Brown, C R

    Dreelin2025CSP APACrefauthors Dreelin, R A. , Brown, C R. \ Hill, J G. APACrefauthors \ 2025 . Keystone Effects of Prairie Dogs (Cynomys spp.) on Grassland Birds: Current Knowledge and Future Directions Keystone effects of prairie dogs (Cynomys spp.) on grassland birds: Current knowledge and future directions . Conservation Science and Practice 7 e70004 ....

    work page doi:10.1111/csp2.70004 2025

  16. [16]

    du2024crosslayer APACrefauthors Du, Z. , Hu, Z. , Zhao, G. , Jin, Y. \ Ma, H. APACrefauthors \ 2024 . Cross-Layer Feature Pyramid Transformer for Small Object Detection in Aerial Images. Cross-layer feature pyramid transformer for small object detection in aerial images

    2024

  17. [17]

    , Biggins, D E

    Eads2011Jmamm APACrefauthors Eads, D A. , Biggins, D E. , Livieri, T M. , Dullum, J L D. , Jachowski, D S. \ Millspaugh, J J. APACrefauthors \ 2011 . Postbreeding Resource Selection by Adult Black-Footed Ferrets in the Conata Basin, South Dakota Postbreeding resource selection by adult black-footed ferrets in the conata basin, south dakota . Journal of Ma...

    work page doi:10.1644/10-mamm-a-250.1 2011

  18. [18]

    , Wind, J

    eikelboom2019improving APACrefauthors Eikelboom, J A. , Wind, J. , Van de Ven, E. , Kenana, L M. , Schroder, B. , De Knegt, H J. Prins, H H. APACrefauthors \ 2019 . Improving the precision and accuracy of animal population estimates with aerial image object detection Improving the precision and accuracy of animal population estimates with aerial image obj...

    2019

  19. [19]

    \ Dale, M R T

    Fortin2005Book APACrefauthors Fortin, M J. \ Dale, M R T. APACrefauthors \ 2005 . Spatial Analysis: A Guide for Ecologists Spatial analysis: A guide for ecologists . Cambridge, UK Cambridge University Press

    2005

  20. [20]

    , Mousavian, A

    georgakis2017synthesizing APACrefauthors Georgakis, G. , Mousavian, A. , Berg, A C. \ Kosecka, J. APACrefauthors \ 2017 . Synthesizing Training Data for Object Detection in Indoor Scenes. Synthesizing training data for object detection in indoor scenes

    2017

  21. [21]

    , Cui, Y

    ghiasi2021copypaste APACrefauthors Ghiasi, G. , Cui, Y. , Srinivas, A. , Qian, R. , Lin, T Y. , Cubuk, E D. Zoph, B. APACrefauthors \ 2021 . Simple Copy-Paste Is a Strong Data Augmentation Method for Instance Segmentation. Simple copy-paste is a strong data augmentation method for instance segmentation. APACrefURL https://openaccess.thecvf.com/content/CVP...

    2021

  22. [22]

    , Sun, X

    Gui2024RSODReview APACrefauthors Gui, S. , Sun, X. , Shi, Y. , Yang, S. , Wang, K. , Hu, W. Zhu, X X. APACrefauthors \ 2024 . Remote Sensing Object Detection in the Deep Learning Era—A Review Remote sensing object detection in the deep learning era—a review . Remote Sensing 16 2 327 . APACrefDOI doi:10.3390/rs16020327 APACrefDOI

    work page doi:10.3390/rs16020327 2024

  23. [23]

    Revisiting pre-trained remote sensing model benchmarks: Resizing and normalization matters

    GuirguisEWKM0B22 APACrefauthors Guirguis, K. , Eskandar, G. , Wang, M. , Kayser, M. , Monari, E. , Yang, B. \ Beyerer, J. APACrefauthors \ 2024 . Uncertainty-based Forgetting Mitigation for Generalized Few-Shot Object Detection. Uncertainty-based forgetting mitigation for generalized few-shot object detection. IEEE . APACrefURL http://dblp.uni-trier.de/db...

    work page doi:10.1109/cvprw63382.2024.00265 2024

  24. [24]

    , Mott, R

    Hodgson2018MEE APACrefauthors Hodgson, J C. , Mott, R. , Baylis, S M. , Pham, T T. , Wotherspoon, S. , Kilpatrick, A D. Koh, L P. APACrefauthors \ 2018 . Drones count wildlife more accurately and precisely than humans Drones count wildlife more accurately and precisely than humans . Methods in Ecology and Evolution 9 5 1160--1167 . APACrefDOI doi:10.1111/...

    work page doi:10.1111/2041-210x.12974 2018

  25. [25]

    APACrefauthors \ 1995

    Hoogland1995Book APACrefauthors Hoogland, J L. APACrefauthors \ 1995 . The Black-Tailed Prairie Dog: Social Life of a Burrowing Mammal The black-tailed prairie dog: Social life of a burrowing mammal . Chicago University of Chicago Press

    1995

  26. [26]

    , Antoniou, A

    hospedales2021meta APACrefauthors Hospedales, T. , Antoniou, A. , Micaelli, P. \ Storkey, A. APACrefauthors \ 2021 . Meta-learning in neural networks: A survey Meta-learning in neural networks: A survey . IEEE transactions on pattern analysis and machine intelligence 44 9 5149--5169

    2021

  27. [27]

    , Smith, J

    Huwaterfowl2024 APACrefauthors Hu, Q. , Smith, J. , Woldt, W. \ Tang, Z. APACrefauthors \ 2021 . UAV-derived waterfowl thermal imagery dataset”, Mendeley Data Uav-derived waterfowl thermal imagery dataset”, mendeley data . Mendeley Data, V4 . APACrefDOI doi:10.17632/46k66mz9sz.4 APACrefDOI

    work page doi:10.17632/46k66mz9sz.4 2021

  28. [28]

    , Tang, X

    huang2022small APACrefauthors Huang, H. , Tang, X. , Wen, F. \ Jin, X. APACrefauthors \ 2022 . Small object detection method with shallow feature fusion network for chip surface defect detection Small object detection method with shallow feature fusion network for chip surface defect detection . Scientific reports 12 1 3914

    2022

  29. [29]

    , Jung, G B

    huynh2024one APACrefauthors Huynh, X N. , Jung, G B. \ Suhr, J K. APACrefauthors \ 2024 . One-Stage Small Object Detection Using Super-Resolved Feature Map for Edge Devices One-stage small object detection using super-resolved feature map for edge devices . Electronics 13 2 409

    2024

  30. [30]

    , Dinesh, C

    kasarla2019region APACrefauthors Kasarla, N. , Dinesh, C. , Tondewad, R. , Radhakrishnan, V B. \ Jawahar, C V. APACrefauthors \ 2019 . Region-Based Active Learning for Efficient Annotation of Semantic Segmentation Datasets. Region-based active learning for efficient annotation of semantic segmentation datasets

    2019

  31. [31]

    , Porensky, L M

    Kearney APACrefauthors Kearney, S P. , Porensky, L M. , Augustine, D J. \ Pellatz, D W. APACrefauthors \ 2023 . Toward broad-scale mapping and characterization of prairie dog colonies from airborne imagery using deep learning Toward broad-scale mapping and characterization of prairie dog colonies from airborne imagery using deep learning . Ecological Indi...

    2023

  32. [32]

    , Marcos, D

    kellenberger2018detecting APACrefauthors Kellenberger, B. , Marcos, D. \ Tuia, D. APACrefauthors \ 2018 . Detecting mammals in UAV images: Best practices to address a substantially imbalanced dataset with deep learning Detecting mammals in uav images: Best practices to address a substantially imbalanced dataset with deep learning . Remote sensing of envir...

    2018

  33. [33]

    , Reading, R P

    Kotliar2000KeystoneConcept APACrefauthors Kotliar, N B. APACrefauthors \ 2000 . Application of the New Keystone-Species Concept to Prairie Dogs: How Well Does It Work? Application of the new keystone-species concept to prairie dogs: How well does it work? Conservation Biology 14 6 1715--1721 . APACrefDOI doi:10.1111/j.1523-1739.2000.98384.x APACrefDOI

    work page doi:10.1111/j.1523-1739.2000.98384.x 2000

  34. [34]

    , Baker, B W

    Kotliar1999CriticalReview APACrefauthors Kotliar, N B. , Baker, B W. , Whicker, A D. \ Plumb, G. APACrefauthors \ 1999 . A Critical Review of Assumptions About the Prairie Dog as a Keystone Species A critical review of assumptions about the prairie dog as a keystone species . Environmental Management 24 2 177--192 . APACrefURL https://pubmed.ncbi.nlm.nih....

    work page doi:10.1007/s002679900013 1999

  35. [35]

    \ Jr., J F C

    KretzerCully2001SWNat APACrefauthors Kretzer, J E. \ Jr., J F C. APACrefauthors \ 2001 . Effects of Black-tailed Prairie Dogs on Reptiles and Amphibians in Kansas Shortgrass Prairie Effects of black-tailed prairie dogs on reptiles and amphibians in kansas shortgrass prairie . The Southwestern Naturalist 46 2 171--177 . APACrefDOI doi:10.2307/3672525 APACrefDOI

    work page doi:10.2307/3672525 2001

  36. [36]

    , Zhang, B

    kumar2024wildlifemapper APACrefauthors Kumar, S. , Zhang, B. , Gudavalli, C. , Levenson, C. , Hughey, L. , Stabach, J A. Manjunath, B. APACrefauthors \ 2024 06 . WildlifeMapper: Aerial Image Analysis for Multi-Species Detection and Identification. Wildlifemapper: Aerial image analysis for multi-species detection and identification

    2024

  37. [37]

    , Babanezhad, R

    laradji2022sealnet APACrefauthors Laradji, I. , Babanezhad, R. , Manfredonia, R. , Vazquez, D. , Nowrouzezahrai, D. , Barrett, L. \ Derpanis, K G. APACrefauthors \ 2022 . SealNet: A Benchmark for Seal Detection and Counting in Aerial Images Sealnet: A benchmark for seal detection and counting in aerial images . Ecological Informatics 67 101527

    2022

  38. [38]

    APACrefauthors \ 1993

    Legendre1993Ecology APACrefauthors Legendre, P. APACrefauthors \ 1993 . Spatial Autocorrelation: Trouble or New Paradigm? Spatial autocorrelation: Trouble or new paradigm? Ecology 74 6 1659--1673 . APACrefDOI doi:10.2307/1939924 APACrefDOI

    work page doi:10.2307/1939924 1993

  39. [39]

    , Wang, Z

    li2024survey APACrefauthors Li, D. , Wang, Z. , Chen, Y. , Jiang, R. , Ding, W. \ Okumura, M. APACrefauthors \ 2024 . A Survey on Deep Active Learning: Recent Advances and New Frontiers. A survey on deep active learning: Recent advances and new frontiers

    2024

  40. [40]

    , Wang, Y

    li2023improved APACrefauthors Li, K. , Wang, Y. \ Hu, Z. APACrefauthors \ 2023 . Improved YOLOv7 for small object detection algorithm based on attention and dynamic convolution Improved yolov7 for small object detection algorithm based on attention and dynamic convolution . Applied Sciences 13 16 9316

    2023

  41. [41]

    , Wei, Y

    li2024YOLOUAV APACrefauthors Li, X. , Wei, Y. , Li, J. , Duan, W. , Zhang, X. \ Huang, Y. APACrefauthors \ 2024 . Improved YOLOv7 algorithm for small object detection in unmanned aerial vehicle image scenarios Improved yolov7 algorithm for small object detection in unmanned aerial vehicle image scenarios . Applied Sciences 14 4 1664

    2024

  42. [42]

    , Doll \'a r, P

    lin2017feature APACrefauthors Lin, T. , Doll \'a r, P. , Girshick, R. , He, K. , Hariharan, B. \ Belongie, S. APACrefauthors \ 2017 07 . Feature Pyramid Networks for Object Detection. Feature pyramid networks for object detection

    2017

  43. [43]

    , Maire, M

    lin2014microsoft APACrefauthors Lin, T. , Maire, M. , Belongie, S. , Hays, J. , Perona, P. , Ramanan, D. Zitnick, C L. APACrefauthors \ 2014 . Microsoft COCO : Common Objects in Context Microsoft COCO : Common objects in context . D. Fleet, T. Pajdla, B. Schiele \ T. Tuytelaars\ ( ), Computer Vision -- ECCV 2014 Computer vision -- eccv 2014 \ ( \ 8693, \ ...

    2014

  44. [44]

    , Huang, B

    rs17071155 APACrefauthors Liu, P. , Huang, B. , Jin, T. \ Long, H. APACrefauthors \ 2025 . Low-Shot Weakly Supervised Object Detection for Remote Sensing Images via Part Domination-Based Active Learning and Enhanced Fine-Tuning Low-shot weakly supervised object detection for remote sensing images via part domination-based active learning and enhanced fine...

    work page doi:10.3390/rs17071155 2025

  45. [45]

    , Lin, Y

    swin APACrefauthors Liu, Z. , Lin, Y. , Cao, Y. , Hu, H. , Wei, Y. , Zhang, Z. Guo, B. APACrefauthors \ 2021 . Swin Transformer: Hierarchical Vision Transformer using Shifted Windows Swin transformer: Hierarchical vision transformer using shifted windows . ICCV

    2021

  46. [46]

    , Lenz, P

    mackowiak2018cereals APACrefauthors Mackowiak, R. , Lenz, P. , Ghori, O. , Diego, F. , Lange, J. \ Rother, C. APACrefauthors \ 2018 . CEREALS—Cost-Effective REgion-based Active Learning for Semantic Segmentation. Cereals—cost-effective region-based active learning for semantic segmentation

    2018

  47. [47]

    , Zhang, B

    mcever2023context APACrefauthors McEver, R A. , Zhang, B. , Levenson, C. , Iftekhar, A. \ Manjunath, B. APACrefauthors \ 2023 . Context-driven detection of invertebrate species in deep-sea video Context-driven detection of invertebrate species in deep-sea video . International Journal of Computer Vision 131 6 1367--1388

    2023

  48. [48]

    , Reading, R P

    Miller2000KeystoneResponse APACrefauthors Miller, B. , Reading, R P. , Hoogland, J L. , Clark, T W. \ Ceballos, G. APACrefauthors \ 2000 . The Role of Prairie Dogs as a Keystone Species: Response to Stapp The role of prairie dogs as a keystone species: Response to stapp . Conservation Biology 14 1 318--321 . APACrefDOI doi:10.1046/j.1523-1739.2000.99201.x...

    work page doi:10.1046/j.1523-1739.2000.99201.x 2000

  49. [49]

    , Liu, T

    mou2023waid APACrefauthors Mou, C. , Liu, T. , Zhu, C. \ Cui, X. APACrefauthors \ 2023 . Waid: A large-scale dataset for wildlife detection with drones Waid: A large-scale dataset for wildlife detection with drones . Applied Sciences 13 18 10397

    2023

  50. [50]

    \ Joubert, D

    Naude_2019_CVPR_Workshops APACrefauthors Naud \'e , J. \ Joubert, D. APACrefauthors \ 2019 06 . The Aerial Elephant Dataset: A New Public Benchmark for Aerial Object Detection. The aerial elephant dataset: A new public benchmark for aerial object detection

    2019

  51. [51]

    YOLOv3: An Incremental Improvement

    redmon2018yolov3 APACrefauthors Redmon, J. \ Farhadi, A. APACrefauthors \ 2018 . YOLOv3: An Incremental Improvement Yolov3: An incremental improvement . arXiv preprint arXiv:1804.02767

    work page internal anchor Pith review arXiv 2018

  52. [52]

    faster_rcnn APACrefauthors Ren, S. , He, K. , Girshick, R B. \ Sun, J. APACrefauthors \ 2015 . Faster R-CNN : Towards Real-Time Object Detection with Region Proposal Networks Faster R-CNN : Towards real-time object detection with region proposal networks . IEEE Transactions on Pattern Analysis and Machine Intelligence

    2015

  53. [53]

    \ Todorovic, S

    roy2018deep APACrefauthors Roy, A. \ Todorovic, S. APACrefauthors \ 2018 . Deep Active Learning for Object Detection. Deep active learning for object detection

    2018

  54. [54]

    APACrefauthors \ 2009

    Settles2009TR APACrefauthors Settles, B. APACrefauthors \ 2009 . Active Learning Literature Survey Active learning literature survey \ \ \ Computer Sciences Technical Report 1648 . University of Wisconsin--Madison . APACrefURL http://burrsettles.com/pub/settles.activelearning.pdf APACrefURL

    2009

  55. [55]

    Khoshgoftaar, A survey on Image Data Augmentation for Deep Learning, J Big Data, 6 (2019), pp

    shorten2019survey APACrefauthors Shorten, C. \ Khoshgoftaar, T M. APACrefauthors \ 2019 . A Survey on Image Data Augmentation for Deep Learning A survey on image data augmentation for deep learning . Journal of Big Data 6 1 60 . APACrefDOI doi:10.1186/s40537-019-0197-0 APACrefDOI

    work page doi:10.1186/s40537-019-0197-0 2019

  56. [56]

    , Valentin, J

    siddiqui2020viewal APACrefauthors Siddiqui, Y. , Valentin, J. \ Nie ner, M. APACrefauthors \ 2020 . ViewAL: Active Learning with Viewpoint Entropy for 3D Object Detection. Viewal: Active learning with viewpoint entropy for 3d object detection

    2020

  57. [57]

    DINOv3

    simeoni2025dinov3 APACrefauthors Sim \'e oni, O. , Vo, H V. , Seitzer, M. , Baldassarre, F. , Oquab, M. , Jose, C. others APACrefauthors \ 2025 . Dinov3 Dinov3 . arXiv preprint arXiv:2508.10104

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  58. [58]

    \ Lomolino, M V

    SmithLomolino2004Oecologia APACrefauthors Smith, G A. \ Lomolino, M V. APACrefauthors \ 2004 . Black-tailed Prairie Dogs and the Structure of Avian Communities on the Shortgrass Plains Black-tailed prairie dogs and the structure of avian communities on the shortgrass plains . Oecologia 138 4 592--602 . APACrefDOI doi:10.1007/s00442-003-1465-3 APACrefDOI

    work page doi:10.1007/s00442-003-1465-3 2004

  59. [59]

    APACrefauthors \ 2007

    Stapp2007Jmamm APACrefauthors Stapp, P. APACrefauthors \ 2007 . Rodent Communities in Active and Inactive Colonies of Black-Tailed Prairie Dogs in Shortgrass Steppe Rodent communities in active and inactive colonies of black-tailed prairie dogs in shortgrass steppe . Journal of Mammalogy 88 1 241--249 . APACrefDOI doi:10.1644/05-MAMM-A-206R3.1 APACrefDOI

    work page doi:10.1644/05-mamm-a-206r3.1 2007

  60. [60]

    , Pang, R

    tan2020efficientdet APACrefauthors Tan, M. , Pang, R. \ Le, Q V. APACrefauthors \ 2020 06 . EfficientDet: Scalable and Efficient Object Detection. Efficientdet: Scalable and efficient object detection

    2020

  61. [61]

    , Camps-Valls, G

    Tuia2011TGRS APACrefauthors Tuia, D. , Camps-Valls, G. , Matasci, G. , Kanevski, M. \ Benediktsson , J A C. APACrefauthors \ 2011 . Active Learning Methods for Remote Sensing Image Classification Active learning methods for remote sensing image classification . IEEE Transactions on Geoscience and Remote Sensing 49 3 868--880 . APACrefDOI doi:10.1109/TGRS....

    work page doi:10.1109/tgrs.2010.2076849 2011

  62. [62]

    , Kellenberger, B

    tuia2022perspectives APACrefauthors Tuia, D. , Kellenberger, B. , Beery, S. , Costelloe, B R. , Zuffi, S. , Risse, B. others APACrefauthors \ 2022 . Perspectives in machine learning for wildlife conservation Perspectives in machine learning for wildlife conservation . Nature communications 13 1 792

    2022

  63. [63]

    , Ratle, F

    tuia2009active APACrefauthors Tuia, D. , Ratle, F. , Pacifici, F. , Kanevski, M F. \ Emery, W J. APACrefauthors \ 2009 . Active Learning Methods for Remote Sensing Image Classification Active learning methods for remote sensing image classification \ ( 47)\ ( 7). APACrefDOI doi:10.1109/TGRS.2008.2010404 APACrefDOI

    work page doi:10.1109/tgrs.2008.2010404 2009

  64. [64]

    , Kretzer, J

    VanNimwegen2008Ecology APACrefauthors VanNimwegen, R E. , Kretzer, J. \ Jr., J F C. APACrefauthors \ 2008 . Ecosystem Engineering by a Colonial Mammal: How Prairie Dogs Structure Rodent Communities Ecosystem engineering by a colonial mammal: How prairie dogs structure rodent communities . Ecology 89 12 3298--3305 . APACrefDOI doi:10.1890/07-1511.1 APACrefDOI

    work page doi:10.1890/07-1511.1 2008

  65. [65]

    , Namkoong, H

    volpi2018generalizing APACrefauthors Volpi, R. , Namkoong, H. , Sener, O. , Duchi, J. , Murino, V. \ Savarese, S. APACrefauthors \ 2018 . Generalizing to Unseen Domains via Adversarial Data Augmentation Generalizing to unseen domains via adversarial data augmentation \ ( 31)

    2018

  66. [66]

    YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors

    wang2022yolov7 APACrefauthors Wang, C Y. , Bochkovskiy, A. \ Liao, H y. APACrefauthors \ 2022 . YOLOv7: Trainable bagof-freebies sets new state-of-the-art for real-time object detectors. 10.48550 Yolov7: Trainable bagof-freebies sets new state-of-the-art for real-time object detectors. 10.48550 . arXiv preprint arXiv.2207.02696

    work page arXiv 2022

  67. [67]

    , Alyami, A

    Yamani_2024_WACV APACrefauthors Yamani, A. , Alyami, A. , Luqman, H. , Ghanem, B. \ Giancola, S. APACrefauthors \ 2024 January . Active Learning for Single-Stage Object Detection in UAV Images. Active learning for single-stage object detection in uav images

    2024

  68. [68]

    , Han, D

    yun2019cutmix APACrefauthors Yun, S. , Han, D. , Oh, S J. , Chun, S. , Choe, J. \ Yoo, Y. APACrefauthors \ 2019 . CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features. Cutmix: Regularization strategy to train strong classifiers with localizable features

    2019

  69. [69]

    , Boulerice, J T

    rarespot_dataset APACrefauthors Zhang, B. , Boulerice, J T. , Kuniyil, N. , Mendiratta, C. , Kumar, S. , Shamon, H. \ Manjunath, B S. APACrefauthors \ 2025 1 . RareSpot+ Prairie Dog Dataset . RareSpot+ Prairie Dog Dataset . https://bisque2.ece.ucsb.edu/client_service/view?resource=https://bisque2.ece.ucsb.edu/data_service/00-B2mLaCeFPkCgZF4LND9uEd . Hoste...

    2025

  70. [70]

    , Boulerice, J T

    zhang2025rarespotspottingsmallrare APACrefauthors Zhang, B. , Boulerice, J T. , Kuniyil, N. , Mendiratta, C. , Kumar, S. , Shamon, H. \ Manjunath, B S. APACrefauthors \ 2025 2 . RareSpot: Spotting Small and Rare Wildlife in Aerial Imagery with Multi-Scale Consistency and Context-Aware Augmentation. Rarespot: Spotting small and rare wildlife in aerial imag...

    2025

  71. [71]

    , Wang, Z

    Zhang_2024 APACrefauthors Zhang, B. , Wang, Z. \ Du, B. APACrefauthors \ 2024 . Boosting Semi-Supervised Object Detection in Remote Sensing Images With Active Teaching Boosting semi-supervised object detection in remote sensing images with active teaching . IEEE Geoscience and Remote Sensing Letters 21 1–5 . APACrefURL http://dx.doi.org/10.1109/LGRS.2024....

    work page doi:10.1109/lgrs.2024.3357098 2024

  72. [72]

    , Ciss \'e , M

    zhang2018mixup APACrefauthors Zhang, H. , Ciss \'e , M. , Dauphin, Y N. \ Lopez - Paz, D. APACrefauthors \ 2018 04 . mixup : Beyond Empirical Risk Minimization. mixup : Beyond empirical risk minimization. International Conference on Learning Representations (ICLR) . APACrefURL https://openreview.net/forum?id=r1Ddp1-Rb APACrefURL OpenReview

    2018

  73. [73]

    , Zhang, H

    zhang2025sodetr APACrefauthors Zhang, H. , Zhang, H. , Mei, A. , Gan, Z. \ Zhu, G N. APACrefauthors \ 2025 . SO-DETR: Leveraging Dual-Domain Features and Knowledge Distillation for Small Object Detection. So-detr: Leveraging dual-domain features and knowledge distillation for small object detection

    2025

  74. [74]

    , Lyu, S

    zhu2021tph APACrefauthors Zhu, X. , Lyu, S. , Wang, X. \ Zhao, Q. APACrefauthors \ 2021 10 . TPH- YOLOv5 : Improved YOLOv5 Based on Transformer Prediction Head for Object Detection on Drone-Captured Scenarios. Tph- YOLOv5 : Improved YOLOv5 based on transformer prediction head for object detection on drone-captured scenarios

    2021

  75. [75]

    , Song, G

    zong2023detrs APACrefauthors Zong, Z. , Song, G. \ Liu, Y. APACrefauthors \ 2023 . Detrs with collaborative hybrid assignments training. Detrs with collaborative hybrid assignments training

    2023