Pith. sign in

REVIEW 4 major objections 6 minor 44 references

Wildlife Target Re-Identification Using Self-supervised Learning in Non-Urban Settings

T0 review · 4 major / 6 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 This paper claims that self-supervised models trained on unlabeled camera-trap pairs outperform supervised baselines for wildlife re-identification across open-world and transfer tasks.

desk verdict Temporal-pair SSL for wildlife re-ID is a plausible idea, but the headline comparison is confounded by dataset differences, and the ablation has an internal contradiction. read the letter →

arxiv 2507.02403 v1 pith:SID67KV5 submitted 2025-07-03 cs.CV cs.AIcs.LG

Mufhumudzi Muthivhi , Terence L. van Zyl This is my paper · ORCID
classification cs.CVcs.AIcs.LG
keywords wildlifere-identificationself-supervisedlearningcameratrapstemporalimagepairsopen-worldtransferrepresentationanimalidentification
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 claims that self-supervised learning trained on automatically extracted temporal image pairs from camera trap footage produces better wildlife re-identification features than models trained with individual class labels. The authors detect animals with an off-the-shelf detector, pair detections across frames at least 120 seconds apart using a 0.2 IoU threshold, and train six SSL methods (SimCLR, MoCo, BarlowTwins, BYOL, FastSiam, DINO) on those pairs. They report that SSL representations beat supervised baselines (Triplet, SupCon, ArcFace) on open-world re-identification, including unseen species and out-of-distribution datasets, and on downstream tasks such as classification, detection, segmentation, attributes, and pose. They also report that SSL models reach competitive accuracy with far fewer training samples, which matters because annotating individual animals at scale is expensive.

What carries the argument

The load-bearing mechanism is the temporal image-pair extractor. A detector finds animal bounding boxes in each camera-trap frame; any detection in a later frame at least 120 seconds away whose box has Intersection over Union (IoU) of at least 0.2 with the original box is treated as a second view of the same individual. These pairs are fed as positive views into standard SSL objectives — contrastive (SimCLR, MoCo), self-distillation (BYOL, DINO, FastSiam), and redundancy reduction (BarlowTwins) — so natural video variation replaces manual augmentation. An ablation shows that lowering the IoU threshold (yielding more pairs) improves mAP and that adding self-distillation on top of temporal pairs helps further, indicating the strategy's effect comes from the volume of true positive temporal views.

What would settle it

Manually label a random sample of the automatically generated temporal pairs and measure the true same-individual rate; if a substantial fraction are different animals, the training signal is noisy and the claimed superiority would likely vanish. Alternatively, train the same SSL methods on temporally shuffled (non-corresponding) pairs and show that the open-world mAP does not drop, which would falsify the claim that temporal correspondence is what drives the gains.

Watch

Extended reading notes

Core claim

The central discovery asserted is that label-free temporal-pair pretraining yields representations that generalize better than class-labeled training for wildlife re-identification. In the paper's in-distribution open-world test, self-supervised variants (SimCLR, BYOL, DINO) reach average mAP of 36-40, at or above the best supervised ArcFace baseline at 36, while Triplet and SupCon lag at 24-27. On out-of-distribution species, SimCLR reaches the highest mAP at 25, ahead of general-purpose pretrained baselines and far ahead of supervised wildlife models (16 or below). Across downstream tasks, DINO and BYOL lead or tie on image classification, video classification, detection, segmentation, and out-of-domain flowers, though TransReID retains the best pose keypoint score. The authors attribute the robustness to temporal pairing, which teaches invariance to pose, lighting, and background variation from real video rather than hand-crafted augmentations.

Load-bearing premise

The load-bearing assumption is that two detections separated by at least 120 seconds and overlapping by at least 20 percent really are the same individual; if this pairing is often wrong, the self-supervised training signal is corrupted and the reported gains would not hold. The paper itself notes that a single detection can be paired with multiple animals and that occlusion or animals entering and exiting frames can cause identity drift.

Editorial extensions

If this is right

  • Wildlife re-identification models can be pretrained from raw camera-trap video with no individual-level annotations, removing the major labeling bottleneck in the field.
  • The same frozen SSL backbone transfers to classification, detection, segmentation, attribute, and pose tasks, suggesting one pretrained encoder can serve multiple conservation analytics pipelines.
  • SSL models stay accurate with fewer training samples, which could make re-identification feasible for rare or data-poor species.
  • Pairing temporal views with self-distillation increases usable training signal, indicating a route to scale with endlessly streaming camera-trap data.
  • The open-world results suggest SSL features are better suited than supervised ones to deployment where new species and individuals appear after deployment.

Reading between the lines

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

  • Beyond the paper's experiments, the temporal-pairing recipe (detector + IoU matching over a time gap) should transfer to other video domains where individuals persist across frames, like livestock or pedestrian tracking, although the 120-second and 0.2 IoU settings would need per-domain tuning.
  • The paper does not test whether the advantage survives when the detector makes errors; a useful extension would be to corrupt or remove detector boxes and measure how mAP degrades, isolating the contribution of pair quality.
  • The finding that dual-encoder SSL methods (BYOL, MoCo, DINO) do best on video classification suggests temporal-pair pretraining may implicitly learn motion or persistence cues; this could be tested by comparing against SSL trained on random frames from the same videos.
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

4 major / 6 minor

Summary. The paper proposes a self-supervised learning pipeline for wildlife re-identification in non-urban settings. The method extracts two views of the same individual from camera trap video by detecting animals with MegaDetector and matching detections in frames at least 120 seconds apart with an IoU threshold of 0.2. Six SSL methods (SimCLR, MoCo, BarlowTwins, BYOL, FastSiam, DINO) are pre-trained on private camera trap data from two South African safaris and evaluated against supervised baselines (Triplet, SupCon, ArcFace) pre-trained on WildlifeReID-10k/wildlife-toolkit data. Evaluation covers open-world re-identification, downstream classification, detection, segmentation, attributes, pose, and out-of-domain transfer. The paper's central claim is that self-supervised models are more robust with limited data and that self-supervised features outperform supervised features across all downstream tasks.

Significance. If the claims were established, the temporal-pair SSL strategy would provide a scalable, annotation-free route to wildlife re-identification and would constitute evidence for SSL transfer advantages in ecological applications. The paper's strengths include a broad comparison across six SSL methods and multiple downstream tasks, an ablation over IoU thresholds, qualitative latent-space and attention-map analyses, and released code. However, the central comparison is confounded by differences in pre-training data between supervised and self-supervised models, so the headline claim is not supported by the reported experiments.

major comments (4)
  1. [Section IV-B, Tables I-III] The central comparison is confounded by training-data differences. Supervised models are pre-trained on the labeled WildlifeReID-10k/wildlife-toolkit data, while all SSL models are pre-trained on private camera trap video from two South African safaris. Because the learning paradigm and the training-data domain, curation, and size vary simultaneously, the observed performance differences cannot be attributed to self-supervision versus supervision. This directly affects the abstract's claim that SSL features outperform supervision across all downstream tasks.
  2. [Table I] Table I labels supervised results as in-distribution and self-supervised results as out-of-distribution. Comparing in-distribution supervised models with out-of-distribution SSL models is not a fair test of open-world re-identification, because the SSL models have not been exposed to the evaluation species or even the same camera-trap domain. A controlled comparison requires evaluating both paradigms on held-out splits drawn from the same data source.
  3. [Figure 1, Section V-A] The 'limited data' argument in Figure 1 is built on a regression over heterogeneous models trained on different datasets, with no error bars. The x-axis mixes labeled WildlifeReID samples and private camera-trap samples, so the plot cannot separate the effect of dataset size from the effect of learning paradigm or data curation. The claim that SSL models are more robust with limited data is therefore not established.
  4. [Section III-B, Section IV-B] The temporal-pair construction assumes that a detection in a later frame with IoU at least 0.2 corresponds to the same individual, and the authors acknowledge in Section IV-B that one detection may be paired with multiple animals and that identity drift can occur. If a substantial fraction of pairs are incorrect, the SSL training signal is corrupted, and the reported gains may not transfer. The manuscript provides no validation of pairing precision or an analysis of how label noise affects the results, so this assumption remains a load-bearing risk for the proposed method.
minor comments (6)
  1. [Abstract] The phrase 'self-supervised learning Self-Supervised Learning (SSL)' contains a duplicated term; it should read 'self-supervised learning (SSL)'.
  2. [Section IV-C] The heading 'Evalution Metrics' is a typo and should be 'Evaluation Metrics'.
  3. [Figure 4] The figure legend says 'Self Distiliation' instead of 'Self Distillation'.
  4. [Table II and related text] The model name 'TransRec' in the Table II caption is inconsistent with 'TransReID' used elsewhere in the paper.
  5. [Figure 1] The x-axis label 'Total Training Samples' is ambiguous: it should specify whether samples are images, crops, or individual identities, and the data sources should be distinguished.
  6. [Abstract] The code URL 'https://github.com/pxpana/' appears truncated; the abstract also refers to 'https://github.com/pxpana/SSLWildlife', so the final link should be corrected and verified.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central SSL-vs-supervised comparison is an empirical evaluation against external benchmarks, not a derivation that reduces to its inputs.

full rationale

The paper's derivation chain is: extract temporal pairs from camera-trap frames using MegaDetector and an IoU threshold (Section III-B), pre-train SSL models on those pairs, and then evaluate the frozen representations on external open-world and transfer benchmarks (Tables I-III). The evaluation labels come from WildlifeReID-10k, iNat21, CIFAR, Animal Kingdom, OxfordPets, COCO, AWA, Plants, and Insects, none of which are used to fit or select the SSL models. The SSL evaluation is therefore not a renaming of the training objective, and the reported mAP, accuracy, IoU, mIoU, multi-label accuracy, and PCK values are empirical results rather than consequences of construction. The temporal-pair identity assumption (IoU >= 0.2 with frames at least 120 seconds apart) is an experimental assumption about positive pairs, and the authors explicitly acknowledge in Section IV-B that it can produce multiple matches or identity drift; this affects the quality of the training signal but does not define the measured outcome. The only self-citations are [3] (a background reference on siamese networks for wildlife identification) and [44] (a future-work pointer on continual learning); neither is load-bearing for the central claim. The comparison does vary both training data and learning paradigm, and supervised variants are hyperparameter-tuned while SSL variants are not, which is a validity concern about whether the gains are attributable to self-supervision per se, but it is not circularity: the numbers are not forced by construction or by a self-citation chain. No circular step was found.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The method introduces no new theoretical entities. It relies on standard SSL losses and a heuristic temporal-pair sampler. The central comparison, however, is built on the unvalidated premise that models trained on different datasets, with different species and image statistics, can be compared to isolate the effect of the learning paradigm.

free parameters (5)
  • IoU threshold alpha = 0.2
    Chosen to maximize the number of temporal pairs; the ablation in V-C1 shows performance varies with this value.
  • Temporal interval between paired frames = 120 seconds
    Arbitrary assumption that individuals leave the camera scene within two minutes; affects the number and quality of positive pairs.
  • MegaDetector confidence threshold = 0.5
    Detections above 0.5 confidence are kept; not varied in the study.
  • KNN k for downstream evaluation = 200
    Fixed following prior work [9]; affects the downstream accuracy values in linear probing and KNN results.
  • Number of SSL training epochs = 100
    Taken from the lightly defaults; no tuning was performed.
assumptions (5)
  • domain assumption MegaDetector produces reliable animal detections in camera trap frames.
    Used in Section III-B and IV-B to extract animal crops from raw video; if detections are poor, temporal pairs are invalid.
  • domain assumption An IoU of at least 0.2 between detections in consecutive frames identifies the same individual.
    Stated in Section III-B; the authors acknowledge in IV-B that this can match one image to multiple animals or lose identity due to occlusion.
  • domain assumption Frames separated by at least 120 seconds but with IoU overlap belong to the same individual and identity drift is negligible.
    Stated in Section IV-B: subsequent frames at least 120 seconds apart are collected assuming the individual would have left the scene after two minutes.
  • domain assumption Standard SSL losses and augmentations are valid for wildlife camera trap imagery without modification.
    Section III-C adopts SimCLR, MoCo, BarlowTwins, BYOL, FastSiam and DINO as-is; no domain-specific adaption is argued.
  • domain assumption The WildlifeReID-10k wildlife subset is a suitable proxy for supervised wildlife re-identification performance.
    Section IV-B uses this dataset to train all supervised baselines; the choice of a public, curated dataset versus private safari video is the central confound.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Wildlife Target Re-Identification Using Self-supervised Learning in Non-Urban Settings." pith.science (2026). https://pith.science/paper/SID67KV5

@misc{pith2026250702403,
  author       = {Pith},
  title        = {Pith review of: Wildlife Target Re-Identification Using Self-supervised Learning in Non-Urban Settings},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SID67KV5}},
  note         = {Machine review of arXiv:2507.02403}
}
read the original abstract

Wildlife re-identification aims to match individuals of the same species across different observations. Current state-of-the-art (SOTA) models rely on class labels to train supervised models for individual classification. This dependence on annotated data has driven the curation of numerous large-scale wildlife datasets. This study investigates self-supervised learning Self-Supervised Learning (SSL) for wildlife re-identification. We automatically extract two distinct views of an individual using temporal image pairs from camera trap data without supervision. The image pairs train a self-supervised model from a potentially endless stream of video data. We evaluate the learnt representations against supervised features on open-world scenarios and transfer learning in various wildlife downstream tasks. The analysis of the experimental results shows that self-supervised models are more robust even with limited data. Moreover, self-supervised features outperform supervision across all downstream tasks. The code is available here https://github.com/pxpana/SSLWildlife.

Figures

Figures reproduced from arXiv: 2507.02403 by the authors.

Figure 1
Figure 1. Comparison of Supervised (Purple) and Self-Supervised (Blue) [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. An illustration of extracting two image pairs from consecutive camera frames. The animals are detected with a bounding box around them then we [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. A sample of temporal pairs, where each pair is highlighted with the [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: mAP achieved by SimCLR across increasing IoU thresholds. Lower [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Latent space separability of ArcFace and SimCLR on open-world data. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Attention maps of models on images within natural environments. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 33 canonical work pages

  1. [1]

    Beyond observation: Deep learning for animal behavior and ecological con- servation,

    L. S. Saoud, A. Sultan, M. Elmezain, M. Heshmat, L. Seneviratne, and I. Hussain, “Beyond observation: Deep learning for animal behavior and ecological con- servation,” Ecological Informatics, p. 102893, 2024

    work page 2024

  2. [2]

    Hotspotter—patterned species instance recognition,

    J. P. Crall, C. V . Stewart, T. Y . Berger-Wolf, D. I. Ruben- stein, and S. R. Sundaresan, “Hotspotter—patterned species instance recognition,” in 2013 IEEE workshop on applications of computer vision (WACV) . IEEE, 2013, pp. 230–237

    work page 2013

  3. [3]

    Automated identification of individuals in wildlife population using siamese neural networks,

    N. Dlamini and T. L. van Zyl, “Automated identification of individuals in wildlife population using siamese neural networks,” in 2020 7th international conference on soft computing & machine intelligence (ISCMI). IEEE, 2020, pp. 224–228

    work page 2020

  4. [4]

    Wildfusion: Individual animal identifica- tion with calibrated similarity fusion,

    V . Cermak, L. Picek, L. Adam, L. Neumann, and J. Matas, “Wildfusion: Individual animal identifica- tion with calibrated similarity fusion,” arXiv preprint arXiv:2408.12934, 2024

    arXiv 2024

  5. [5]

    Toward re-identifying any animal,

    B. Jiao, L. Liu, L. Gao, R. Wu, G. Lin, P. Wang, and Y . Zhang, “Toward re-identifying any animal,”Advances in Neural Information Processing Systems , vol. 36, pp. 40 042–40 053, 2023

    work page 2023

  6. [6]

    Wildlifedatasets: An open-source toolkit for animal re- identification,

    V . ˇCerm´ak, L. Picek, L. Adam, and K. Papafitsoros, “Wildlifedatasets: An open-source toolkit for animal re- identification,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , 2024, pp. 5953–5963

    work page 2024

  7. [7]

    WildlifeReID-10k: Wildlife re-identification dataset with 10k individual animals

    L. Adam, V . ˇCerm´ak, K. Papafitsoros, and L. Picek, “Wildlifereid-10k: Wildlife re-identification dataset with 10k individual animals,” arXiv preprint arXiv:2406.09211, 2024

    work page Pith review arXiv 2024

  8. [8]

    Multispecies animal re-id using a large community-curated dataset,

    L. Otarashvili, T. Subramanian, J. Holmberg, J. Lev- enson, and C. V . Stewart, “Multispecies animal re-id using a large community-curated dataset,” arXiv preprint arXiv:2412.05602, 2024

    arXiv 2024

Show all 44 references
  1. [9]

    Unsupervised feature learning via non-parametric instance discrimina- tion,

    Z. Wu, Y . Xiong, S. X. Yu, and D. Lin, “Unsupervised feature learning via non-parametric instance discrimina- tion,” in Proceedings of the IEEE conference on com- puter vision and pattern recognition , 2018, pp. 3733– 3742

    2018

  2. [10]

    Distinctive image features from scale- invariant keypoints,

    D. G. Lowe, “Distinctive image features from scale- invariant keypoints,” International journal of computer vision, vol. 60, pp. 91–110, 2004

    2004

  3. [11]

    Su- perpoint: Self-supervised interest point detection and description,

    D. DeTone, T. Malisiewicz, and A. Rabinovich, “Su- perpoint: Self-supervised interest point detection and description,” in Proceedings of the IEEE conference on computer vision and pattern recognition workshops , 2018, pp. 224–236

    2018

  4. [12]

    Bioclip: A vision foundation model for the tree of life,

    S. Stevens, J. Wu, M. J. Thompson, E. G. Campolongo, C. H. Song, D. E. Carlyn, L. Dong, W. M. Dahdul, C. Stewart, T. Berger-Wolf et al. , “Bioclip: A vision foundation model for the tree of life,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recogni...

    2024

  5. [13]

    Transreid: Transformer-based object re-identification,

    S. He, H. Luo, P. Wang, F. Wang, H. Li, and W. Jiang, “Transreid: Transformer-based object re-identification,” in Proceedings of the IEEE/CVF international conference on computer vision , 2021, pp. 15 013–15 022

    2021

  6. [14]

    Un- derstanding the impact of training set size on animal re- identification,

    A. Algasov, E. Nepovinnykh, T. Eerola, H. K ¨alvi¨ainen, C. V . Stewart, L. Otarashvili, and J. A. Holmberg, “Un- derstanding the impact of training set size on animal re- identification,” arXiv preprint arXiv:2405.15976 , 2024

    2024 arXiv

  7. [15]

    replicant: a pipeline for generating anno- tated images of animals in complex environments using unreal engine,

    F. Plum, R. Bulla, H. K. Beck, N. Imirzian, and D. Labonte, “replicant: a pipeline for generating anno- tated images of animals in complex environments using unreal engine,” Nature Communications, vol. 14, no. 1, p. 7195, 2023

    2023

  8. [16]

    A critical analysis of self-supervision, or what we can learn from a single image,

    Y . M. Asano, C. Rupprecht, and A. Vedaldi, “A critical analysis of self-supervision, or what we can learn from a single image,” arXiv preprint arXiv:1904.13132, 2019

    1904 arXiv

  9. [17]

    A simple framework for contrastive learning of visual representations,

    T. Chen, S. Kornblith, M. Norouzi, and G. Hinton, “A simple framework for contrastive learning of visual representations,” in International conference on machine learning. PmLR, 2020, pp. 1597–1607

    2020

  10. [18]

    Mo- mentum contrast for unsupervised visual representation learning,

    K. He, H. Fan, Y . Wu, S. Xie, and R. Girshick, “Mo- mentum contrast for unsupervised visual representation learning,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 9729–9738

    2020

  11. [19]

    Exploring simple siamese rep- resentation learning,

    X. Chen and K. He, “Exploring simple siamese rep- resentation learning,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2021, pp. 15 750–15 758

    2021

  12. [20]

    Bootstrap your own latent- a new approach to self-supervised learning,

    J.-B. Grill, F. Strub, F. Altch ´e, C. Tallec, P. Richemond, E. Buchatskaya, C. Doersch, B. Avila Pires, Z. Guo, M. Gheshlaghi Azar et al., “Bootstrap your own latent- a new approach to self-supervised learning,” Advances in neural information processing systems , vol. 33, pp....

    2020

  13. [21]

    Barlow twins: Self-supervised learning via redundancy reduction,

    J. Zbontar, L. Jing, I. Misra, Y . LeCun, and S. Deny, “Barlow twins: Self-supervised learning via redundancy reduction,” in International conference on machine learn- ing. PMLR, 2021, pp. 12 310–12 320

    2021

  14. [22]

    Emerging properties in self-supervised vision transformers,

    M. Caron, H. Touvron, I. Misra, H. J ´egou, J. Mairal, P. Bojanowski, and A. Joulin, “Emerging properties in self-supervised vision transformers,” in Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 9650–9660

    2021

  15. [23]

    Unsupervised learning of visual features by contrasting cluster assignments,

    M. Caron, I. Misra, J. Mairal, P. Goyal, P. Bojanowski, and A. Joulin, “Unsupervised learning of visual features by contrasting cluster assignments,” Advances in neural information processing systems, vol. 33, pp. 9912–9924, 2020

    2020

  16. [24]

    Sim- ilarity of neural network representations revisited,

    S. Kornblith, M. Norouzi, H. Lee, and G. Hinton, “Sim- ilarity of neural network representations revisited,” in International conference on machine learning . PMLR, 2019, pp. 3519–3529

    2019

  17. [25]

    Towards out-of-distribution generalization: A survey,

    J. Liu, Z. Shen, Y . He, X. Zhang, R. Xu, H. Yu, and P. Cui, “Towards out-of-distribution generalization: A survey,” arXiv preprint arXiv:2108.13624 , 2021

    2021 arXiv

  18. [26]

    Self-supervised features improve open-world learning,

    A. R. Dhamija, T. Ahmad, J. Schwan, M. Jafarzadeh, C. Li, and T. E. Boult, “Self-supervised features improve open-world learning,” arXiv preprint arXiv:2102.07848 , 2021

    2021 arXiv

  19. [27]

    How well do self-supervised models transfer?

    L. Ericsson, H. Gouk, and T. M. Hospedales, “How well do self-supervised models transfer?” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 5414–5423

    2021

  20. [28]

    Efficient pipeline for camera trap image review,

    S. Beery, D. Morris, and S. Yang, “Efficient pipeline for camera trap image review,” 2019

    2019

  21. [29]

    Decoupled contrastive learning,

    C.-H. Yeh, C.-Y . Hong, Y .-C. Hsu, T.-L. Liu, Y . Chen, and Y . LeCun, “Decoupled contrastive learning,” in Eu- ropean conference on computer vision . Springer, 2022, pp. 668–684

    2022

  22. [30]

    Improved deep metric learning with multi- class n-pair loss objective,

    K. Sohn, “Improved deep metric learning with multi- class n-pair loss objective,” Advances in neural infor- mation processing systems , vol. 29, 2016

    2016

  23. [31]

    Fast- siam: Resource-efficient self-supervised learning on a single gpu,

    D. Pototzky, A. Sultan, and L. Schmidt-Thieme, “Fast- siam: Resource-efficient self-supervised learning on a single gpu,” in DAGM German Conference on Pattern Recognition. Springer, 2022, pp. 53–67

    2022

  24. [32]

    Arcface: Additive angular margin loss for deep face recognition,

    J. Deng, J. Guo, N. Xue, and S. Zafeiriou, “Arcface: Additive angular margin loss for deep face recognition,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2019, pp. 4690–4699

    2019

  25. [33]

    In defense of the triplet loss for person re-identification,

    A. Hermans, L. Beyer, and B. Leibe, “In defense of the triplet loss for person re-identification,” arXiv preprint arXiv:1703.07737, 2017

    2017 arXiv

  26. [34]

    Su- pervised contrastive learning,

    P. Khosla, P. Teterwak, C. Wang, A. Sarna, Y . Tian, P. Isola, A. Maschinot, C. Liu, and D. Krishnan, “Su- pervised contrastive learning,” Advances in neural infor- mation processing systems , vol. 33, pp. 18 661–18 673, 2020

    2020

  27. [35]

    Lightly

    I. Susmelj, M. Heller, P. Wirth, J. Prescott, M. Ebner, and et al., “Lightly.” [Online]. Available: https: //github.com/lightly-ai/lightly

  28. [36]

    Benchmarking representation learn- ing for natural world image collections,

    G. Van Horn, E. Cole, S. Beery, K. Wilber, S. Belongie, and O. Mac Aodha, “Benchmarking representation learn- ing for natural world image collections,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 12 884–12 893

    2021

  29. [37]

    Learning multiple layers of features from tiny images.(2009),

    A. Krizhevsky, G. Hinton et al. , “Learning multiple layers of features from tiny images.(2009),” 2009

    2009

  30. [38]

    Animal kingdom: A large and diverse dataset for animal behavior understanding,

    X. L. Ng, K. E. Ong, Q. Zheng, Y . Ni, S. Y . Yeo, and J. Liu, “Animal kingdom: A large and diverse dataset for animal behavior understanding,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 19 023–19 034

    2022

  31. [39]

    Animals detection images dataset,

    A. Jana, “Animals detection images dataset,” Kaggle, n.d., accessed: October 2023. [On- line]. Available: https://www.kaggle.com/datasets/ antoreepjana/animals-detection-images-dataset/data

    2023

  32. [40]

    Microsoft coco: Common objects in context,

    T.-Y . Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Doll ´ar, and C. L. Zitnick, “Microsoft coco: Common objects in context,” in Computer vision– ECCV 2014: 13th European conference, zurich, Switzer- land, September 6-12, 2014, proceedings, part v 13 . Spring...

    2014

  33. [41]

    Cats and dogs,

    O. M. Parkhi, A. Vedaldi, A. Zisserman, and C. V . Jawa- har, “Cats and dogs,” in IEEE Conference on Computer Vision and Pattern Recognition , 2012

    2012

  34. [42]

    Zero- shot learning—a comprehensive evaluation of the good, the bad and the ugly,

    Y . Xian, C. H. Lampert, B. Schiele, and Z. Akata, “Zero- shot learning—a comprehensive evaluation of the good, the bad and the ugly,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 41, no. 9, pp. 2251–2265, 2018

    2018

  35. [43]

    Automated flower classification over a large number of classes,

    M.-E. Nilsback and A. Zisserman, “Automated flower classification over a large number of classes,” in 2008 Sixth Indian conference on computer vision, graphics & image processing. IEEE, 2008, pp. 722–729

    2008

  36. [44]

    Incremental class learning using variational autoencoders with similarity learning,

    J. Huo and T. L. van Zyl, “Incremental class learning using variational autoencoders with similarity learning,” Neural Computing and Applications , vol. 37, no. 2, pp. 769–784, 2025

    2025

Pith tools

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