Pith. sign in

REVIEW 4 major objections 5 minor 33 references

PictSure: Pretraining Embeddings Matters for In-Context Learning Image Classifiers

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read In visual in-context learning, the pretrained embedding space, not the fusion transformer's training data, determines few-shot accuracy.

desk verdict PictSure's positive claim about encoder quality holds up; its negative claim about fusion-training diversity needs seeds and a stricter mixture before it carries. read the letter →

arxiv 2506.14842 v2 pith:CYJXDK6E submitted 2025-06-16 cs.CV cs.AI

Lukas Schiesser , Cornelius Wolff , Sophie Haas , Simon Pukrop This is my paper
classification cs.CVcs.AI
keywords in-contextlearningfew-shotimageclassificationpretrainedembeddingsrepresentationqualityfusiontransformerdomaingeneralizationDINOv2CLIP
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

PictSure is a vision-only in-context image classifier: a frozen pretrained encoder produces embeddings for labeled support images and a query, and a small fusion transformer reads those tokens and predicts the query label with no parameter updates at inference time. The paper's central claim is that on this pipeline the quality of the encoder's pretrained embeddings is the dominant driver of few-shot accuracy, while broadening the fusion transformer's training data from ImageNet alone to a sixteen-dataset multi-domain mixture brings little or no gain. This matters because it redirects research effort toward representation learning rather than fusion-module complexity or task diversity. If correct, it means simpler ICL systems built on strong embeddings can match or beat more elaborate few-shot methods under domain shift.

What carries the argument

The load-bearing object is the frozen pretrained encoder feeding a small in-context fusion transformer (four blocks, eight attention heads, model dimension 1024) through concatenated support-label and query tokens under an asymmetric attention mask: support tokens see each other, and the query attends to all support tokens, but support tokens cannot attend to the query. This design isolates the encoder as the only variable that changes between model variants, so observed differences in 5-way 1-shot and 5-shot accuracy are attributed to the embedding space. The paper's argument is that the fusion transformer learns to read embeddings from ImageNet-21K episodes alone and transfers that reading ability to new domains, so further gains must come from the encoder's pretraining objective and data.

What would settle it

Fix the fusion transformer and compare two encoders with identical architecture and parameter count but different pretraining objectives and data scales (for example, the same ViT trained self-supervised on general images versus supervised on ImageNet); if ICL accuracy does not track the measured class-separability of the embeddings, or if a multi-domain fusion training set restores performance for the weaker encoder, the central claim would be falsified.

Watch

Extended reading notes

Core claim

Varying only the frozen encoder while keeping the fusion architecture fixed, the authors find that DINOv2 and CLIP embeddings consistently outperform supervised ResNet embeddings on tieredImageNet, PlantDoc, Brain Tumor, and OCTMNIST, in both 1-shot and 5-shot settings. Training the fusion transformer on a deliberately diverse mixture of sixteen datasets instead of ImageNet-21K produces nearly identical results, with the only notable exception being the ResNet backbone, whose weaker embeddings give the fusion layer more to gain from varied training tasks. The paper concludes that the limitation of visual in-context learning is representation quality, not the fusion transformer's exposure to diverse tasks, and that an ICL transformer behaves as a robust reader of whatever embedding space it is given.

Load-bearing premise

The load-bearing premise is that the three encoders differ mainly in pretraining quality, when in fact ResNet, CLIP, and DINOv2 also differ in architecture, capacity, dataset size, and training compute, so attributing performance differences to representation quality is not fully isolated.

Editorial extensions

If this is right

  • If the bottleneck is representation quality, improving encoder pretraining objectives and data should yield larger few-shot in-context learning gains than enlarging or diversifying fusion-transformer training sets.
  • Vision-only ICL models with strong frozen embeddings, such as DINOv2, can be competitive with or beat language-grounded CLIP pipelines like CAML on out-of-domain medical benchmarks.
  • The fusion-layer training recipe can stay simple: ImageNet-21K episodes suffice, and multi-domain mixtures add little when the encoder is strong.
  • The released open weights and MCP server let agentic systems call few-shot image classification directly inside AI pipelines, making the encoder-centric result immediately usable.
  • A supervised MLP trained on the same frozen embeddings often matches or beats the more elaborate baselines, reinforcing that pretrained embeddings already contain substantial task-relevant information.

Reading between the lines

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

  • Editorial inference: the encoder comparison conflates architecture, capacity, and pretraining data, so the practical claim should be read as 'the encoder matters more than fusion training diversity' rather than 'self-supervision per se is best'.
  • Editorial inference: if the fusion transformer is a general reader of embedding spaces, the same PictSure fusion module could be attached to future encoders without retraining, and its few-shot accuracy should track that encoder's embedding quality.
  • Editorial inference: the ResNet exception hints at a threshold effect: below a certain embedding quality, fusion training diversity starts to matter, so the central claim may hold only above that threshold.
  • Editorial inference: the negligible gain from multi-domain fusion training suggests that few-shot meta-learning research should shift competition from fusion complexity to representation-learning benchmarks.
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 / 5 minor

Summary. The paper introduces PictSure, a vision-only in-context learning (ICL) architecture with a frozen pretrained encoder followed by a small fusion transformer that reads support-query token sequences under an asymmetric mask. The authors claim two central findings: (i) representation quality induced by encoder pretraining strongly correlates with downstream few-shot image classification performance, and (ii) varying the fusion transformer's training data from ImageNet-21K alone to a diverse 16-dataset mixture yields limited additional gains, implying that the bottleneck in visual ICL is representation quality rather than fusion-module training diversity. The method is evaluated on tieredImageNet, PlantDoc, Brain Tumor, and OCTMNIST against RFS, PMF, CAML, Qwen3.5, and a supervised MLP probe on embeddings. The authors release model weights and an MCP server for agentic integration.

Significance. If the central claims hold, the paper provides a useful and potentially influential result for the visual ICL community: it would redirect attention from fusion-layer design and training-data diversity toward encoder representation quality. The paper is commendable for releasing open-source weights and a tooling interface, for evaluating across in-domain and out-of-domain benchmarks including medical and agricultural imagery, and for including an MLP probe that independently tests embedding quality. The main contribution is empirical rather than theoretical, and the conclusions are actionable for future work on pretraining objectives and encoder architectures.

major comments (4)
  1. [Section 3, Figure 2] The negative claim that 'varying the training dataset for the fusion transformer ... provides limited additional performance gains' rests on a comparison between the ImageNet and Default fusion-training settings, but no training-seed variance is reported. The standard errors shown in Figure 2 are over 2000 evaluation episodes, not over independent training runs, so the near-identical results could be a Type II error, especially given the small fusion transformer (four blocks, dimension 1024) that may have saturated its capacity. Please report results across at least three to five training seeds per condition, or provide a statistical test that accounts for training-run variability.
  2. [Section 3, Default mixture] The Default mixture assigns 20% of training episodes to ILSVRC-2012, which is a subset of the ImageNet-21K class set used in the ImageNet condition. The two fusion-training conditions therefore share a substantial portion of their task distribution, which weakens the diversity manipulation. The claim that fusion-module training diversity is not the bottleneck would be much more convincing with a strict non-ImageNet mixture that excludes ILSVRC-2012 and other ImageNet-derived data, or with an analysis that controls for the overlap.
  3. [Section 3, encoder comparisons] The three encoders compared (ResNet, CLIP, DINOv2) differ simultaneously in architecture (CNN vs ViT), model capacity, pretraining dataset size, and pretraining objective. The statement that 'representation quality induced by pretraining strongly correlates with downstream ICL performance' is therefore not isolated from architecture and scale confounds. Please add controlled comparisons, such as the same ViT architecture pretrained with different objectives, or otherwise temper the causal attribution in Sections 5 and 6.
  4. [Section 3, convergence claim] The claim that 'models without pretrained encoders fail to converge' is stated in the introduction and Section 3 but is not supported by any reported experiment, figure, or table. If this claim is load-bearing for the centrality of pretraining, please provide convergence curves or final training accuracy for an untrained-encoder baseline. As written, the reader cannot verify this assertion.
minor comments (5)
  1. [Section 3 vs Introduction] The description of ResNet pretraining is inconsistent: the introduction says ResNet is 'pretrained on ImageNet-21k', while Section 3 says it is 'trained on ImageNet1K'. Please clarify which dataset and resolution were used.
  2. [Figure 2] The caption and legend of Figure 2 are difficult to parse; the distinction between the ImageNet and Default fusion-training conditions is not clearly legible, especially in black-and-white. Please restructure the figure or caption to make the compared conditions explicit.
  3. [Section 4 vs Figure 2] Section 4 states that all evaluations use 5-way tasks, but Figure 2 labels BrainTumor and OCTMNIST as '4-way'. Please correct this inconsistency in the text or the figure.
  4. [Abstract and contributions] The abstract and contribution list mention 'OrganCMNIST', but the experiments evaluate 'OCTMNIST'. Please unify the dataset name.
  5. [References] Several bibliography entries contain line breaks inside URLs (e.g., the Kaggle and arXiv links), which may break hyperlinks in the published version. Please ensure all URLs are correctly formatted.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's central claims are empirical comparisons backed by an independent embedding-quality probe, not derivations from their own outputs.

full rationale

The paper's conclusions are empirical rather than derived. The encoder comparison (ResNet vs. CLIP vs. DINOv2) and the fusion-training comparison (ImageNet vs. Default) are controlled experiments reported in Section 3, Figure 2, and Section 5; no equation defines the target result in terms of the inputs, and no parameter fitted to a subset is later renamed as a prediction. The representation-quality claim is supported by an independent probe: the supervised MLP trained directly on frozen embeddings (Section 4), which is neither the ICL transformer nor constructed from PictSure's outputs. The claim that fusion-module training diversity is not the bottleneck rests on a near-null empirical comparison; its fragility (no reported training-seed variance and a 20% ILSVRC-2012 overlap with the ImageNet-21K condition) is a statistical/design concern, not a circularity. The bibliography contains no self-citations, so no load-bearing self-citation chain exists. The conclusion echoes the external RFS result [26], but that result is cited as prior evidence and tested in a new architecture setting, which is independent support rather than a renaming of the present paper's own finding. No circular step can be exhibited with a quote, so the score is 0.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The paper does not introduce new mathematical entities or physical objects. The free parameters are the hand-chosen fusion-training mixture and the fusion transformer's hyperparameters, which the central null result depends on. The axioms are domain assumptions about encoder transferability and the sufficiency of ImageNet-21K, plus the interpretive assumption that encoder differences reflect representation quality rather than confounded factors.

free parameters (2)
  • Default mixture dataset weights = ILSVRC-2012 20%, Places365 10%, VGGFace2 10%, Products10K 10%, LEGO Parts 10%, Food-101 5%, SD-198 5%, ChestX 5%…
    The conclusion that fusion training data diversity does not matter is tested only with this hand-picked mixture, so the negative result is conditional on this composition and these weights.
  • Fusion transformer hyperparameters = 4 blocks, 8 attention heads, model dimension 1024, 10,000 episodes per epoch
    The fusion module's capacity and training budget are chosen by hand; the claim that the fusion module is not the bottleneck assumes this configuration is representative.
assumptions (3)
  • domain assumption Frozen pretrained image encoders produce embeddings that a small fusion transformer can read without adaptation.
    Used throughout Sections 3 and 5 to interpret performance differences as encoder effects.
  • domain assumption ImageNet-21K is a sufficient fusion-training corpus to teach general support-query interaction for unseen domains.
    The central negative result depends on this; a less diverse or smaller fusion corpus might have shown gains from the multi-domain mixture.
  • ad hoc to paper Differences among ResNet, CLIP, and DINOv2 are attributable to representation quality rather than architecture, scale, or compute.
    This is the interpretive leap in Section 5; the experiment does not control for these confounds.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PictSure: Pretraining Embeddings Matters for In-Context Learning Image Classifiers." pith.science (2026). https://pith.science/paper/CYJXDK6E

@misc{pith2026250614842,
  author       = {Pith},
  title        = {Pith review of: PictSure: Pretraining Embeddings Matters for In-Context Learning Image Classifiers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CYJXDK6E}},
  note         = {Machine review of arXiv:2506.14842}
}
read the original abstract

Building image classification models remains cumbersome in data-scarce domains, where collecting large labeled datasets is impractical. In-context learning (ICL) is a promising paradigm for few-shot image classification (FSIC), but prior work has underexplored the relative importance of encoder pretraining versus fusion-layer training data. We present PictSure, a vision-only ICL family of models that demonstrates the potential of easy-to-use fusion transformer architectures, as well as the need for better embedding representations across a wider range of image domains. In both in-domain and out-of-domain evaluations, we find that representation quality induced by pretraining strongly correlates with downstream ICL performance. Crucially, varying the training dataset for the fusion transformer, from ImageNet alone to diverse multi-domain mixtures, provides limited additional performance gains under the evaluated settings, demonstrating that the fusion layer appears capable of adapting effectively once embeddings are sufficiently structured. These results show that the bottleneck in visual ICL is representation quality, not fusion-module training diversity. To facilitate adoption and reproducibility, we release all model weights as open-source artifacts and provide an MCP server that exposes PictSure as a callable tool for LLM-based agentic systems, enabling few-shot image classification to be invoked directly within AI pipelines without integration overhead. Code can be found at https://github.com/PictSure and models at https://huggingface.co/pictsure.

Figures

Figures reproduced from arXiv: 2506.14842 by the authors.

Figure 1
Figure 1. Overview of the PictSure architecture [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Test results across all PictSure model variants compared to baselines 5 Analysis Encoder quality appears to be one of the strongest factors influencing PictSure’s performance. DINOv2 and CLIP variants consistently outperform the ResNet￾based variant across all benchmarks, most clearly on tieredImageNet and Plant￾Doc in both shot settings. These results support the hypothesis that strong representations may contribut… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

33 extracted references · 20 canonical work pages

  1. [1]

    Lego Parts Classification Dataset - 2.4M Renders (Oct 2025), URLhttps: //www.kaggle.com/datasets/mihahamyt/lego-parts-classificatio n-dataset-2-4m-renders

    work page 2025

  2. [2]

    Adam, L., Čermák, V., Papafitsoros, K., Picek, L.: WildlifeReID-10k: Wildlife re-identification dataset with 10k individual animals (Apr 2025), https://doi.org/10.48550/arXiv.2406.09211, URLhttp://arxiv.or g/abs/2406.09211, arXiv:2406.09211 [cs]

    arXiv doi:10.48550/arxiv.2406.09211 2025

  3. [3]

    Data6(11), 111 (Oct 2021), ISSN 2306-5729,https://doi.org/ 10.3390/data6110111, URLhttps://www.mdpi.com/2306-5729/6/11/1 11

    Alsolami, A.S., Shalash, W., Alsaggaf, W., Ashoor, S., Refaat, H., Elmogy, M.: King Abdulaziz University Breast Cancer Mammogram Dataset (KAU- BCMD). Data6(11), 111 (Oct 2021), ISSN 2306-5729,https://doi.org/ 10.3390/data6110111, URLhttps://www.mdpi.com/2306-5729/6/11/1 11

    work page doi:10.3390/data6110111 2021

  4. [4]

    Bai, Y., Chen, Y., Yu, W., Wang, L., Zhang, W.: Products-10K: A Large- scale Product Recognition Dataset (Aug 2020),https://doi.org/10.4 8550/arXiv.2008.10545, URLhttp://arxiv.org/abs/2008.10545, arXiv:2008.10545 [cs]

    arXiv doi:10.48550/arxiv.2008.10545 2020

  5. [5]

    Bossard, L., Guillaumin, M., Van Gool, L.: Food-101 – Mining Discrimina- tive Components with Random Forests. In: Fleet, D., Pajdla, T., Schiele, B., Tuytelaars,T.(eds.)ComputerVision–ECCV2014,vol.8694,pp.446–461, Springer International Publishing, Cham (2014), ISBN 978-3-319-10598-7 978-3-319-10599-4,https://doi.org/10.1007/978-3-319-10599-4_29, URLhttp:...

    doi:10.1007/978-3-319-10599-4_29 2014

  6. [6]

    In: Advances in Neural Information Processing Systems, vol

    Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J.D., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert- Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D., Wu, J., Winter, C., Hesse, C., Chen, M., Sigler, E., Litwin, M., Gray, S., Chess, B., Clark, J., Berner, C., McCandlish, S., Radford, A.,...

    work page 2020

  7. [7]

    Cao, Q., Shen, L., Xie, W., Parkhi, O.M., Zisserman, A.: VGGFace2: A dataset for recognising faces across pose and age (May 2018),https://do i.org/10.48550/arXiv.1710.08092, URLhttp://arxiv.org/abs/1710 .08092, arXiv:1710.08092 [cs]

    arXiv doi:10.48550/arxiv.1710.08092 2018

  8. [8]

    Fifty, C., Duan, D., Junkins, R.G., Amid, E., Leskovec, J., Re, C., Thrun, S.: Context-Aware Meta-Learning (Oct 2023), URLhttps://openreview .net/forum?id=lJYAkDVnRU

    work page 2023

Show all 33 references
  1. [9]

    In: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp

    He, K., Zhang, X., Ren, S., Sun, J.: Deep Residual Learning for Image Recognition. In: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770–778, IEEE, Las Vegas, NV, USA (Jun 2016), 8 L. Schiesser et al. ISBN 978-1-4673-8851-1,https://doi.org/10.1109...

    2016

  2. [10]

    Helber, P., Bischke, B., Dengel, A., Borth, D.: EuroSAT: A Novel Dataset and Deep Learning Benchmark for Land Use and Land Cover Classification. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing12(7), 2217–2226 (Jul 2019), ISSN 1939-1404, 2151-15...

    2019

  3. [11]

    In: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp

    Hu, S.X., Li, D., Stuhmer, J., Kim, M., Hospedales, T.M.: Pushing the Limits of Simple Pipelines for Few-Shot Learning: External Data and Fine- Tuning Make a Difference. In: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 9058–9067, IEEE, New Or...

    2022

  4. [12]

    In: Proceedings of the IEEE International Conference on Computer Vision Workshops (ICCVW), pp

    Krause, J., Stark, M., Deng, J., Fei-Fei, L.: 3D Object Representations for Fine-Grained Categorization. In: Proceedings of the IEEE International Conference on Computer Vision Workshops (ICCVW), pp. 554–561, Syd- ney, Australia (2013),https://doi.org/10.1109/ICCVW.2013.77, UR...

    2013

  5. [13]

    Nickparvar, M.: Brain tumor mri dataset (2026),https://doi.org/10.3 4740/KAGGLE/DSV/14832123, URLhttps://www.kaggle.com/dsv/14832 123

    2026

  6. [14]

    In: 2008 Sixth Indian Conference on Computer Vision, Graphics & Image Processing, pp

    Nilsback, M.E., Zisserman, A.: Automated Flower Classification over a Large Number of Classes. In: 2008 Sixth Indian Conference on Computer Vision, Graphics & Image Processing, pp. 722–729, IEEE, Bhubaneswar, India (Dec 2008),https://doi.org/10.1109/ICVGIP.2008.47, URL http://...

    2008

  7. [15]

    Oquab, M., Darcet, T., Moutakanni, T., Vo, H.V., Szafraniec, M., Khali- dov, V., Fernandez, P., Haziza, D., Massa, F., El-Nouby, A., Assran, M., Ballas, N., Galuba, W., Howes, R., Huang, P.Y., Li, S.W., Misra, I., Rab- bat, M., Sharma, V., Synnaeve, G., Xu, H., Jegou, H., Mair...

    2024

  8. [16]

    IEEE Transactions on Multimedia26, 3469–3480 (2024), ISSN 1520-9210, 1941-0077,https://doi.org/10.1109/TMM.2023.331 1646, URLhttps://ieeexplore.ieee.org/document/10243119/

    Peng, F., Yang, X., Xiao, L., Wang, Y., Xu, C.: SgVA-CLIP: Semantic- Guided Visual Adapting of Vision-Language Models for Few-Shot Image Classification. IEEE Transactions on Multimedia26, 3469–3480 (2024), ISSN 1520-9210, 1941-0077,https://doi.org/10.1109/TMM.2023.331 1646, UR...

    2024

  9. [17]

    In: International confer- ence on machine learning, pp

    Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., others: Learning transferable visual models from natural language supervision. In: International confer- ence on machine learning, pp. 8748–8763, PmLR (2021)

    2021

  10. [18]

    Ren, M., Triantafillou, E., Ravi, S., Snell, J., Swersky, K., Tenenbaum, J.B., Larochelle, H., Zemel, R.S.: META-LEARNING FOR SEMI-SUPERVISED FEW-SHOT CLASSIFICATION (2018) PictSure: Pretraining Embeddings for ICL Image Classifiers 9

    2018

  11. [19]

    In: Vanschoren, J., Yeung, S

    Ridnik, T., Ben-Baruch, E., Noy, A., Zelnik, L.: ImageNet-21K Pretraining for the Masses. In: Vanschoren, J., Yeung, S. (eds.) Proceedings of the Neu- ral Information Processing Systems Track on Datasets and Benchmarks, vol. 1 (2021), URLhttps://datasets-benchmarks-proceedings...

    2021

  12. [20]

    Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., Berg, A.C., Fei-Fei, L.: Im- ageNet Large Scale Visual Recognition Challenge. International Journal of Computer Vision115(3), 211–252 (Dec 2015), ISSN 0920-...

    2015 doi

  13. [21]

    S, S., Dharani Devi, G., V, R., Jeyalakshmi, J.: Privacy-Preserving Breast Cancer Classification: A Federated Transfer Learning Approach. Journal of Imaging Informatics in Medicine37(4), 1488–1504 (Feb 2024), ISSN 2948- 2933,https://doi.org/10.1007/s10278-024-01035-8, URLhttps...

    2024 doi

  14. [22]

    sensors16(8), 1222 (2016)

    Sa, I., Ge, Z., Dayoub, F., Upcroft, B., Perez, T., McCool, C.: Deepfruits: A fruit detection system using deep neural networks. sensors16(8), 1222 (2016)

    2016

  15. [23]

    In: Proceedings of the 7th ACM IKDD CoDS and 25th COMAD, pp

    Singh, D., Jain, N., Jain, P., Kayal, P., Kumawat, S., Batra, N.: PlantDoc: A Dataset for Visual Plant Disease Detection. In: Proceedings of the 7th ACM IKDD CoDS and 25th COMAD, pp. 249–253, ACM, Hyderabad India (Jan 2020), ISBN 978-1-4503-7738-6,https://doi.org/10.1145/33711...

    2020

  16. [24]

    Advances in neural information processing systems30(2017)

    Snell, J., Swersky, K., Zemel, R.: Prototypical networks for few-shot learn- ing. Advances in neural information processing systems30(2017)

    2017

  17. [25]

    In: European conference on computer vision, pp

    Sun, X., Yang, J., Sun, M., Wang, K.: A benchmark for automatic visual classification of clinical skin disease images. In: European conference on computer vision, pp. 206–222, Springer (2016)

    2016

  18. [26]

    266–282, Springer (2020)

    Tian, Y., Wang, Y., Krishnan, D., Tenenbaum, J.B., Isola, P.: Rethinking few-shot image classification: a good embedding is all you need? In: Com- puterVision–ECCV2020:16thEuropeanConference,Glasgow,UK,August 23–28, 2020, Proceedings, Part XIV 16, pp. 266–282, Springer (2020)

    2020

  19. [27]

    Scientific Data5(1), 180161 (Aug 2018), ISSN 2052-4463,https: //doi.org/10.1038/sdata.2018.161, URLhttps://www.nature.com/a rticles/sdata2018161

    Tschandl, P., Rosendahl, C., Kittler, H.: The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions. Scientific Data5(1), 180161 (Aug 2018), ISSN 2052-4463,https: //doi.org/10.1038/sdata.2018.161, URLhttps://www.nature.com/a...

    2018 doi

  20. [28]

    In: 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp

    Wang, X., Peng, Y., Lu, L., Lu, Z., Bagheri, M., Summers, R.M.: ChestX- ray8: Hospital-Scale Chest X-Ray Database and Benchmarks on Weakly- Supervised Classification and Localization of Common Thorax Diseases. In: 2017 IEEE Conference on Computer Vision and Pattern Recognition...

    2017

  21. [29]

    Plant Methods18(1), 28 (2022)

    Yang, J., Guo, X., Li, Y., Marinello, F., Ercisli, S., Zhang, Z.: A survey of few-shot learning in smart agriculture: developments, applications, and challenges. Plant Methods18(1), 28 (2022)

    2022

  22. [30]

    Scientific Data10(1), 41 (Jan 2023), ISSN 2052-4463,https://doi.org/10.1038/s41597- 022- 01721-8, URL https://www.nature.com/articles/s41597-022-01721-8

    Yang, J., Shi, R., Wei, D., Liu, Z., Zhao, L., Ke, B., Pfister, H., Ni, B.: MedMNIST v2 - A large-scale lightweight benchmark for 2D and 3D biomedical image classification. Scientific Data10(1), 41 (Jan 2023), ISSN 2052-4463,https://doi.org/10.1038/s41597- 022- 01721-8, URL ht...

    2023 doi

  23. [31]

    Zhang, S., Metaxas, D.: On the challenges and perspectives of foundation modelsformedicalimageanalysis.Medicalimageanalysis91,102996(2024)

    2024

  24. [32]

    Zhou, B., Lapedriza, A., Khosla, A., Oliva, A., Torralba, A.: Places: A 10 Million Image Database for Scene Recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence40(6), 1452–1464 (Jun 2018), ISSN 0162-8828, 2160-9292, 1939-3539,https://doi.org/10.1109/TPAM...

    2018

  25. [33]

    Sensors24(18), 6109 (2024)

    Zhou, Y., Yan, H., Ding, K., Cai, T., Zhang, Y.: Few-shot image classifi- cation of crop diseases based on vision–language models. Sensors24(18), 6109 (2024)

    2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.