Pith. sign in

REVIEW 4 major objections 5 minor 54 references

Component Adaptive Clustering for Generalized Category Discovery

T0 review · 4 major / 5 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 AdaGCD combines adaptive slot attention over DINO local features with global contrastive learning and reports state-of-the-art results on six GCD benchmarks.

desk verdict AdaGCD is a plausible incremental GCD method that combines AdaSlot with CMS-style contrastive learning, but the key ablation is missing, so the slot-based mechanism remains unproven. read the letter →

arxiv 2507.01711 v1 pith:FEJ6LC3C submitted 2025-07-02 cs.CV

Mingfu Yan , Jiancheng Huang , Yifan Liu , Shifeng Chen This is my paper
classification cs.CV
keywords GeneralizedCategoryDiscoveryAdaptiveSlotAttentioncontrastivelearningDINOlocalfeaturesspatialinformationfine-grainedclassificationopen-world
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

The paper proposes AdaGCD, a contrastive learning framework for Generalized Category Discovery (GCD), where a model must label images from both labeled and unseen classes without knowing how many unseen classes there are. The central claim is that the 2D spatial feature maps produced by the DINO backbone carry category-discriminative local structure that global features alone miss, and that decomposing this structure with an Adaptive Slot Attention module improves clustering. AdaSlot chooses how many slots to use for each image through a Gumbel-Softmax selection module, so the number of components is not fixed in advance. The pooled slot features are concatenated with DINO's global feature and trained with contrastive objectives adapted from a mean-shift baseline. On six benchmarks the method reports the highest overall clustering accuracy in its comparison table, with the largest gain on Herbarium 19 (50.1 versus 45.1).

What carries the argument

The load-bearing object is the component clusterer $S_\theta$, built on Adaptive Slot Attention (AdaSlot). Given the DINO local feature map $h \in \mathbb{R}^{H \times W \times D}$ and $K_{\max}$ Gaussian-initialized slots $s_{\mathrm{init}}$, AdaSlot uses a Gumbel-Softmax selection module to choose a per-image number of slots and a masked slot decoder $d_\theta$ to reconstruct $h$ under the loss $\mathcal{L}^{\mathrm{rec}}_\theta = \| h - \hat{h} \|$. The selected output slots are average-pooled and concatenated with the DINO global feature $g_{\mathrm{dino}}$ to form the image-scale vector in $\mathbb{R}^{3D}$ that drives the supervised and unsupervised contrastive losses. The adaptive slot count is what removes the need to pre-specify how many components each image contains.

What would settle it

Ablate the component clusterer by replacing AdaSlot with simple mean pooling of the DINO local feature map (or with fixed-slot Slot Attention) and rerun the six benchmarks; if the all-class accuracies do not fall materially, the claim that adaptive slot components drive the improvement is falsified.

Watch

Extended reading notes

Core claim

Concretely, the paper reports all-class clustering accuracy of 83.4 on CIFAR100, 87.0 on ImageNet100, 71.4 on CUB, 59.6 on Stanford Cars, 56.4 on FGVC Aircraft, and 50.1 on Herbarium 19, each the highest All score in Table I. The method treats an image as an assembly of semantic components: DINO's final-layer local features are passed through AdaSlot, which selects a per-image subset from a maximum of $K_{\max}$ slots and reconstructs the local feature map through a masked decoder under an MSE loss. Average-pooling the selected slot features and concatenating them with the DINO global representation gives an image-scale vector in $\mathbb{R}^{3D}$ that is trained with the supervised and unsupervised contrastive losses borrowed from CMS. The paper's message is that representing images through components rather than a single global summary is what lets the model separate novel classes, especially in fine-grained domains.

Load-bearing premise

The method's gain rests on the assumption that the MSE reconstruction of DINO local features forces the slots to capture category-discriminative components rather than generic texture; if the slots instead encode low-level patterns, the pooled slot features would add noise and the reported improvements would not transfer.

Editorial extensions

If this is right

  • The reported All accuracy on all six benchmarks in Table I is higher than every comparison method, so pooling adaptive slot features with the DINO global feature is a competitive strategy for GCD.
  • The dynamic slot count removes the need to predefine how many components each image contains, although the method still requires a maximum capacity $K_{\max}$ and the evaluation still assumes a known class count via Hungarian matching.
  • The largest gains appear on fine-grained and specialist datasets (CUB, Stanford Cars, FGVC Aircraft, Herbarium 19), where category differences rest on subtle local structure rather than global appearance.
  • New-class accuracy improves while old-class accuracy stays strong, indicating the local-feature branch does not sacrifice known categories for novel ones.
  • Loss-weight tuning trades old-class retention against new-class adaptation; with weights 0.6 unsupervised, 0.3 supervised, and 0.1 reconstruction the paper reports its best balance.

Reading between the lines

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

  • Editorial inference: the same component-clustering recipe should transfer to other self-supervised vision transformers whose final-layer features keep spatial structure, provided their slots align with semantic parts; this is testable by swapping the DINO backbone.
  • Editorial inference: because AdaSlot picks a per-image slot count, the framework is naturally suited to multi-object or multi-label images where a single global feature cannot represent all categories, a regime the paper does not evaluate.
  • Editorial inference: the reconstruction term operates in DINO feature space, so the method inherits DINO's biases; with a less part-aligned backbone the MSE objective would likely need reweighting or replacement, and comparing that behavior would clarify how much of the gain is due to the slot mechanism versus the backbone.
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 proposes AdaGCD, a generalized category discovery method that augments the CMS contrastive objective with Adaptive Slot Attention (AdaSlot). DINO ViT local feature maps are decomposed into a variable number of slots, the slots are pooled into an image-level vector, this vector is concatenated with the DINO global feature, and the combined representation is trained with supervised and unsupervised contrastive losses plus an MSE reconstruction loss. Experiments on six GCD benchmarks report state-of-the-art or competitive accuracy, with the largest gain on Herbarium 19 (50.1 vs. 45.1 for LegoGCD). Ablations vary the loss weights and the maximum slot number Kmax.

Significance. If the reported gains are reproducible and can be attributed to the slot-based local features, AdaGCD would be a useful contribution: it brings object-centric and spatially localized representations into GCD, and its adaptive slot-count mechanism addresses a practical limitation of fixed-capacity slot models. The paper provides full comparison tables on six datasets, loss-ratio ablations, a Kmax sensitivity study, and a t-SNE visualization. However, the central attribution of the gains to the slot mechanism is not isolated by any ablation, and there is a dimensional inconsistency in the feature concatenation, so the significance is conditional on a clarifying revision.

major comments (4)
  1. [Sec. IV-D, Tables III and IV] The central claim that clustered spatial local features drive the performance gains is not established, because no ablation removes the slot branch or replaces it with a simpler pooling operation. Table III varies only the loss weights while keeping the AdaSlot component, and Table IV varies Kmax but all rows still include AdaSlot plus the reconstruction loss. The single fixed-slot row compares only Kmax=5. To isolate the effect of Adaptive Slot Attention, the authors should add ablations such as (a) CMS plus the reconstruction loss without slots, and (b) CMS with mean pooling of DINO patch features concatenated to the global feature. Without these, the reported improvements over CMS could arise from the auxiliary MSE regularization or from simply appending an extra feature stream rather than from slots that capture category-discriminative local semantics.
  2. [Sec. III-B] The text states that the pooled clustered features are concatenated with the global feature gdino ∈ RD and that this "results in a unified vector gall ∈ R3D." Since both the global feature and the pooled slot features are D-dimensional, the concatenation should be in R2D, not R3D. This dimensional mismatch is unexplained and is important for reproducibility; the authors should specify exactly which vectors are concatenated (including whether the two augmented views are combined) and correct the stated dimension.
  3. [Abstract and Sec. III-A] The claim that the method "removes the need for predefined slot counts" is overstated. Adaptive Slot Attention still requires the maximum slot number Kmax as a predefined hyperparameter, and Table IV shows that accuracy varies with Kmax (e.g., All accuracy ranges from 69.2 to 71.4). The manuscript should describe the contribution as adaptively selecting the number of slots up to a maximum, state how the number of active slots is determined (threshold or selection criterion), and discuss the dependence on Kmax in relation to the claimed adaptivity.
  4. [Table I and Sec. IV-C] Several reported margins are small (e.g., ImageNet100 Old: 95.2 vs. 95.6 for CMS; FGVC Aircraft New: 49.4 vs. 52.3 for CMS) and all results appear to come from a single run without standard deviations or multiple seeds. Since the overall SOTA claims depend on these close comparisons, the authors should report mean and standard deviation over at least three seeds, or release code so that the results can be reproduced and variance assessed.
minor comments (5)
  1. [Eq. (1)] The loss is called an MSE loss but is written as ||h − ĥ|| without a square; the exact form (L2 norm vs. squared L2 norm, and whether it is averaged over pixels) should be stated precisely.
  2. [Table II] The table header is confusing: "Classes labeled unlabeled labeled unlabeled" does not clearly indicate that the columns refer to known classes, unknown classes, labeled images, and unlabeled images. Please rename the columns.
  3. [Sec. III-A] The notation "slot-scale features si_out" is used without first defining the index i; the slot indexing and the relationship of si_out to sout ∈ R^{Kmax×D} should be made explicit.
  4. [References] Reference [41] for Gumbel-Softmax is cited as ICLR 2022, but the original Gumbel-Softmax paper appeared at ICLR 2017; please correct the citation or use the appropriate version.
  5. [Fig. 3] The caption says "Visualization of ten instances" but the figure shows t-SNE embeddings of learned representations; the caption should state that the plot is a t-SNE projection of the learned features for sampled instances.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: AdaGCD is an empirical composition of previously published modules (AdaSlot, CMS) and benchmark evaluations; no central claim reduces by construction to its inputs.

full rationale

The paper's derivation chain is not circular. Eq. (1) defines an MSE reconstruction loss for the slot decoder, and Eq. (2) defines the total loss as a weighted combination of that reconstruction loss and the CMS contrastive losses; these are training objectives, not outputs derived from the claimed result. The component clusterer is explicitly borrowed from Adaptive Slot Attention [20], and the image-scale contrastive losses are taken from CMS [26]; citing these prior works is legitimate external support rather than a self-citation loop. The paper's own references include several previous papers by the same group, but none of those citations supplies a load-bearing premise for the GCD claim; they appear in the general deep-learning context. The main empirical comparison is against other published methods on standard benchmarks, so the SOTA claim is externally grounded. The absence of an ablation that isolates slot pooling from the reconstruction regularizer (Sec. IV-D), and the apparent typo in Sec. III-B that concatenating two D-dimensional vectors yields R^{3D}, are limitations, but they are not circular: the reported gains are not equivalent to a fitted parameter or to a definition. No step defines AdaGCD's output in terms of the benchmark accuracy, and no parameter is fitted to a subset and then 'predicted' as the same quantity. Therefore no circularity is present.

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

The empirical claim rests on pretrained DINO features, the CMS loss implementation, and hand-set hyperparameters. The paper introduces no new physical or conceptual entity; the adaptive slot module is borrowed from prior work.

free parameters (3)
  • Kmax (maximum slot count) = 50
    Set to 50 in all main experiments; ablation in Table IV shows accuracy varies with Kmax, so the model still requires this hand-chosen upper bound despite the claim of adaptive slot counts.
  • Loss weights lambda_U_IC, lambda_S_IC, lambda_rec = 0.6, 0.3, 0.1
    Chosen following [1] and used across datasets; Table III shows sensitivity, especially on Stanford Cars and Herbarium, so the result depends on this selection.
  • Clustered feature dimensionality = 64
    Architectural choice for slot features; no analysis of sensitivity is provided, so it remains a hand-set design parameter.
assumptions (3)
  • domain assumption DINO ViT-B/16 final-layer spatial features preserve local semantic information useful for GCD.
    The method depends on decomposing DINO local feature maps into slots; if these features lack local semantics, the component clusterer has nothing useful to extract. Invoked in Sec. III-A.
  • domain assumption MSE reconstruction of DINO features by a masked slot decoder encourages slots to learn category-relevant components.
    Eq. 1 is the only training signal for the slot branch; the paper provides no analysis linking reconstruction quality to clustering performance. Invoked in Sec. III-A.
  • domain assumption Following the CMS [26] implementation of contrastive losses is a valid baseline, and the only meaningful change is the addition of slot features.
    AdaGCD is a modification of CMS; any differences in training procedures could confound the comparison. Invoked in Sec. III-B.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Component Adaptive Clustering for Generalized Category Discovery." pith.science (2026). https://pith.science/paper/FEJ6LC3C

@misc{pith2026250701711,
  author       = {Pith},
  title        = {Pith review of: Component Adaptive Clustering for Generalized Category Discovery},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FEJ6LC3C}},
  note         = {Machine review of arXiv:2507.01711}
}
read the original abstract

Generalized Category Discovery (GCD) tackles the challenging problem of categorizing unlabeled images into both known and novel classes within a partially labeled dataset, without prior knowledge of the number of unknown categories. Traditional methods often rely on rigid assumptions, such as predefining the number of classes, which limits their ability to handle the inherent variability and complexity of real-world data. To address these shortcomings, we propose AdaGCD, a cluster-centric contrastive learning framework that incorporates Adaptive Slot Attention (AdaSlot) into the GCD framework. AdaSlot dynamically determines the optimal number of slots based on data complexity, removing the need for predefined slot counts. This adaptive mechanism facilitates the flexible clustering of unlabeled data into known and novel categories by dynamically allocating representational capacity. By integrating adaptive representation with dynamic slot allocation, our method captures both instance-specific and spatially clustered features, improving class discovery in open-world scenarios. Extensive experiments on public and fine-grained datasets validate the effectiveness of our framework, emphasizing the advantages of leveraging spatial local information for category discovery in unlabeled image datasets.

Figures

Figures reproduced from arXiv: 2507.01711 by the authors.

Figure 1
Figure 1. Illustration of the Adaptive Slot Attention [20]. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. This figure illustrates our proposed method, AdaGCD, a cluster-centric contrastive learning framework for Generalized Category Discovery (GCD). [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Visualization of ten instances across six datasets using t-SNE [42]. Each subfigure corresponds to a different dataset. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

54 extracted references · 46 canonical work pages

  1. [1]

    Generalized category discovery,

    S. Vaze, K. Han, A. Vedaldi, and A. Zisserman, “Generalized category discovery,” in CVPR, 2022

    work page 2022

  2. [2]

    Gpt4motion: Scripting physical motions in text-to-video generation via blender-oriented gpt planning,

    J. Lv, Y . Huang, M. Yan, J. Huang, J. Liu, Y . Liu, Y . Wen, X. Chen, and S. Chen, “Gpt4motion: Scripting physical motions in text-to-video generation via blender-oriented gpt planning,” in CVPR Workshop, 2024

    work page 2024

  3. [3]

    Sbcr: Stochasticity beats content restriction problem in training and tuning free image editing,

    J. Huang, M. Yan, Y . Liu, and S. Chen, “Sbcr: Stochasticity beats content restriction problem in training and tuning free image editing,” in ICMR, 2024

    work page 2024

  4. [4]

    COMAE: COMprehensive Attribute Exploration for Zero-shot Hashing

    Y . Li, Q. Long, Y . Zhou, N. Cao, S. Liu, F. Zheng, Z. Zhu, Z. Ning, M. Xiao, X. Wang, et al., “Comae: Comprehensive attribute exploration for zero-shot hashing,” arXiv preprint arXiv:2402.16424 , 2024

    work page Pith review arXiv 2024

  5. [5]

    Magicfight: Personalized martial arts combat video generation,

    J. Huang, M. Yan, S. Chen, Y . Huang, and S. Chen, “Magicfight: Personalized martial arts combat video generation,” in ACM MM, 2024

    work page 2024

  6. [6]

    Sglp: A similarity guided fast layer partition pruning for compressing large deep models,

    Y . Li, Y . Lu, Z. Dong, C. Yang, Y . Chen, and J. Gou, “Sglp: A similarity guided fast layer partition pruning for compressing large deep models,” arXiv preprint arXiv:2410.14720 , 2024

    arXiv 2024

  7. [7]

    Dive: Taming dino for subject-driven video editing,

    Y . Huang, W. Xiong, H. Zhang, C. Chen, J. Liu, M. Yan, and S. Chen, “Dive: Taming dino for subject-driven video editing,” arXiv preprint arXiv:2412.03347, 2024

    arXiv 2024

  8. [8]

    Federated Knowledge Distillation for Multi-Model Architectures Lithography Hotspot Detection

    Y . Li, X. Lin, K. Zhang, C. Yang, Z. Guo, J. Gou, and Y . Li, “Fedkd- hybrid: Federated hybrid knowledge distillation for lithography hotspot detection,” arXiv preprint arXiv:2501.04066 , 2025

    work page Pith review arXiv 2025

Show all 54 references
  1. [9]

    Color-sd: Stable diffusion model already has a color style noisy latent space,

    J. Huang, M. Yan, Y . Liu, and S. Chen, “Color-sd: Stable diffusion model already has a color style noisy latent space,” in ICME, 2024

    2024

  2. [10]

    Wavedm: Wavelet-based diffusion models for image restoration,

    Y . Huang, J. Huang, J. Liu, M. Yan, Y . Dong, J. Lv, C. Chen, and S. Chen, “Wavedm: Wavelet-based diffusion models for image restoration,” IEEE Transactions on Multimedia , 2024

    2024

  3. [11]

    Enhanc- ing image generation fidelity via progressive prompts,

    Z. Xiong, Y . Li, C. Yang, T. Tan, Z. Zhu, S. Li, and Y . Ma, “Enhanc- ing image generation fidelity via progressive prompts,” arXiv preprint arXiv:2501.07070, 2025

    2025 arXiv

  4. [12]

    Diffusion model-based image editing: A survey,

    Y . Huang, J. Huang, Y . Liu, M. Yan, J. Lv, J. Liu, W. Xiong, H. Zhang, S. Chen, and L. Cao, “Diffusion model-based image editing: A survey,” arXiv preprint arXiv:2402.17525 , 2024

    2024 arXiv

  5. [13]

    Wizardmath: Empowering mathematical rea- soning for large language models via reinforced evol-instruct,

    H. Luo, Q. Sun, C. Xu, P. Zhao, J. Lou, C. Tao, X. Geng, Q. Lin, S. Chen, and D. Zhang, “Wizardmath: Empowering mathematical rea- soning for large language models via reinforced evol-instruct,” arXiv preprint arXiv:2308.09583, 2023

    2023 arXiv

  6. [14]

    Wizardarena: Post-training large language models via simulated offline chatbot arena,

    H. Luo, Q. Sun, C. Xu, P. Zhao, Q. Lin, J.-G. Lou, S. Chen, Y . Tang, and W. Chen, “Wizardarena: Post-training large language models via simulated offline chatbot arena,” Advances in NeurIPS , 2024

    2024

  7. [15]

    Distilling self-supervised vision transformers for weakly-supervised few-shot classification & segmentation,

    D. Kang, P. Koniusz, M. Cho, and N. Murray, “Distilling self-supervised vision transformers for weakly-supervised few-shot classification & segmentation,” in CVPR, 2023

    2023

  8. [16]

    Collab- orative learning of semi-supervised segmentation and classification for medical images,

    Y . Zhou, X. He, L. Huang, L. Liu, F. Zhu, S. Cui, and L. Shao, “Collab- orative learning of semi-supervised segmentation and classification for medical images,” in CVPR, 2019

    2019

  9. [17]

    Meta-learning for semi-supervised few- shot classification,

    M. Ren, E. Triantafillou, S. Ravi, J. Snell, K. Swersky, J. B. Tenenbaum, H. Larochelle, and R. S. Zemel, “Meta-learning for semi-supervised few- shot classification,” in ICLR, 2018

    2018

  10. [18]

    Semi-supervised domain adaptation via sample-to-sample self-distillation,

    J. Yoon, D. Kang, and M. Cho, “Semi-supervised domain adaptation via sample-to-sample self-distillation,” in IEEE Winter Conference on Applications of Computer Vision (WACV) , pp. 1978–1987, 2022

    1978

  11. [19]

    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 ICCV, 2021

    2021

  12. [20]

    Adaptive slot attention: Object discovery with dynamic slot number,

    K. Fan, Z. Bai, T. Xiao, T. He, M. Horn, Y . Fu, F. Locatello, and Z. Zhang, “Adaptive slot attention: Object discovery with dynamic slot number,” in CVPR, pp. 23062–23071, 2024

    2024

  13. [21]

    Learning to discover novel visual categories via deep transfer clustering,

    K. Han, A. Vedaldi, and A. Zisserman, “Learning to discover novel visual categories via deep transfer clustering,” in ICCV, 2019

    2019

  14. [22]

    Dynamic conceptional contrastive learning for generalized category discovery,

    N. Pu, Z. Zhong, and N. Sebe, “Dynamic conceptional contrastive learning for generalized category discovery,” in CVPR, 2023

    2023

  15. [23]

    Promptcal: Contrastive affinity learning via auxiliary prompts for generalized novel category discovery,

    S. Zhang, S. Khan, Z. Shen, M. Naseer, G. Chen, and F. S. Khan, “Promptcal: Contrastive affinity learning via auxiliary prompts for generalized novel category discovery,” in CVPR, pp. 3479–3488, 2023

    2023

  16. [24]

    Parametric classification for generalized category discovery: A baseline study,

    X. Wen, B. Zhao, and X. Qi, “Parametric classification for generalized category discovery: A baseline study,” in ICCV, pp. 16590–16600, 2023

    2023

  17. [25]

    Para- metric information maximization for generalized category discovery,

    F. Chiaroni, J. Dolz, Z. I. Masud, A. Mitiche, and I. Ben Ayed, “Para- metric information maximization for generalized category discovery,” in ICCV, pp. 1729–1739, 2023

    2023

  18. [26]

    Contrastive mean-shift learning for generalized category discovery,

    S. Choi, D. Kang, and M. Cho, “Contrastive mean-shift learning for generalized category discovery,” in CVPR, pp. 23094–23104, 2024

    2024

  19. [27]

    Solving the catastrophic forgetting problem in generalized category discovery,

    X. Cao, X. Zheng, G. Wang, W. Yu, Y . Shen, K. Li, Y . Lu, and Y . Tian, “Solving the catastrophic forgetting problem in generalized category discovery,” in CVPR, pp. 16880–16889, 2024

    2024

  20. [28]

    Unsupervised learning of dense visual representations,

    P. O. O Pinheiro, A. Almahairi, R. Benmalek, F. Golemo, and A. C. Courville, “Unsupervised learning of dense visual representations,” Advances in NeurIPS , vol. 33, pp. 4489–4500, 2020

    2020

  21. [29]

    Dense contrastive learning for self-supervised visual pre-training,

    X. Wang, R. Zhang, C. Shen, T. Kong, and L. Li, “Dense contrastive learning for self-supervised visual pre-training,” in CVPR, 2021

    2021

  22. [30]

    Propagate yourself: Exploring pixel-level consistency for unsupervised visual representation learning,

    Z. Xie, Y . Lin, Z. Zhang, Y . Cao, S. Lin, and H. Hu, “Propagate yourself: Exploring pixel-level consistency for unsupervised visual representation learning,” in CVPR, 2021

    2021

  23. [31]

    Unsupervised semantic segmentation by contrasting object mask pro- posals,

    W. Van Gansbeke, S. Vandenhende, S. Georgoulis, and L. Van Gool, “Unsupervised semantic segmentation by contrasting object mask pro- posals,” in ICCV, pp. 10052–10062, 2021

    2021

  24. [32]

    Unsupervised semantic segmentation with self-supervised object- centric representations,

    A. Zadaianchuk, M. Kleindessner, Y . Zhu, F. Locatello, and T. Brox, “Unsupervised semantic segmentation with self-supervised object- centric representations,” in ICLR, 2023

    2023

  25. [33]

    Casting your model: Learning to localize improves self-supervised representations,

    R. R. Selvaraju, K. Desai, J. Johnson, and N. Naik, “Casting your model: Learning to localize improves self-supervised representations,” in CVPR, pp. 11058–11067, 2021

    2021

  26. [34]

    Efficient visual pretraining with contrastive detection,

    O. J. H ´enaff, S. Koppula, J.-B. Alayrac, A. Van den Oord, O. Vinyals, and J. Carreira, “Efficient visual pretraining with contrastive detection,” in ICCV, pp. 10086–10096, 2021

    2021

  27. [35]

    Self-supervised visual representation learning from hierarchical grouping,

    X. Zhang and M. Maire, “Self-supervised visual representation learning from hierarchical grouping,” Advances in NeurIPS , 2020

    2020

  28. [36]

    Self-supervised visual representation learning with semantic grouping,

    X. Wen, B. Zhao, A. Zheng, X. Zhang, and X. Qi, “Self-supervised visual representation learning with semantic grouping,” in Advances in NeurIPS, 2022

    2022

  29. [37]

    Bridging the gap to real-world object-centric learning,

    M. Seitzer, M. Horn, A. Zadaianchuk, D. Zietlow, T. Xiao, C.-J. Simon- Gabriel, T. He, Z. Zhang, B. Sch ¨olkopf, T. Brox, and F. Locatello, “Bridging the gap to real-world object-centric learning,” in ICLR, 2023

    2023

  30. [38]

    Object discovery and represen- tation networks,

    O. J. H ´enaff, S. Koppula, E. Shelhamer, D. Zoran, A. Jaegle, A. Zisser- man, J. Carreira, and R. Arandjelovi ´c, “Object discovery and represen- tation networks,” in ECCV, Springer, 2022

    2022

  31. [39]

    Object-centric learning with slot attention,

    F. Locatello, D. Weissenborn, T. Unterthiner, A. Mahendran, G. Heigold, J. Uszkoreit, A. Dosovitskiy, and T. Kipf, “Object-centric learning with slot attention,” Advances in NeurIPS , vol. 33, pp. 11525–11538, 2020

    2020

  32. [40]

    Genesis-v2: Inferring unordered object representations without iterative refinement,

    M. Engelcke, O. Parker Jones, and I. Posner, “Genesis-v2: Inferring unordered object representations without iterative refinement,” Advances in NeurIPS , vol. 34, pp. 8085–8094, 2021

    2021

  33. [41]

    Categorical reparameterization with gumbel-softmax,

    E. Jang, S. Gu, and B. Poole, “Categorical reparameterization with gumbel-softmax,” in ICLR, 2022

    2022

  34. [42]

    Visualizing data using t-sne.,

    L. Van der Maaten and G. Hinton, “Visualizing data using t-sne.,” Journal of machine learning research , vol. 9, no. 11, 2008

    2008

  35. [43]

    Learning multiple layers of features from tiny images,

    A. Krizhevsky and G. Hinton, “Learning multiple layers of features from tiny images,” Technical report, 2009

    2009

  36. [44]

    Imagenet: A large-scale hierarchical image database,

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “Imagenet: A large-scale hierarchical image database,” in CVPR, 2009

    2009

  37. [45]

    The caltech-ucsd birds-200-2011 dataset,

    C. Wah, S. Branson, P. Welinder, P. Perona, and S. Belongie, “The caltech-ucsd birds-200-2011 dataset,” Technical Report CNS-TR-2011- 001, California Institute of Technology , 2011

    2011

  38. [46]

    3d object representations for fine-grained categorization,

    J. Krause, M. Stark, J. Deng, and L. Fei-Fei, “3d object representations for fine-grained categorization,” in 4th International IEEE Workshop on 3D Representation and Recognition (3dRR-13) , 2013

    2013

  39. [47]

    Fine- grained visual classification of aircraft,

    S. Maji, E. Rahtu, J. Kannala, M. Blaschko, and A. Vedaldi, “Fine- grained visual classification of aircraft,” arXiv preprint arXiv:1306.5151, 2013

    2013 arXiv

  40. [48]

    The herbarium challenge 2019 dataset,

    K. C. Tan, Y . Liu, B. Ambrose, M. Tulig, and S. Belongie, “The herbarium challenge 2019 dataset,” in Workshop on Fine-Grained Visual Categorization, 2019

    2019

  41. [49]

    Automatically discovering and learning new visual categories with ranking statistics,

    K. Han, S.-A. Rebuffi, S. Ehrhardt, A. Vedaldi, and A. Zisserman, “Automatically discovering and learning new visual categories with ranking statistics,” in ICLR, 2020

    2020

  42. [50]

    A unified objective for novel class discovery,

    E. Fini, E. Sangineto, S. Lathuili `ere, Z. Zhong, M. Nabi, and E. Ricci, “A unified objective for novel class discovery,” in ICCV, 2021

    2021

  43. [51]

    Open-world semi-supervised learning,

    K. Cao, M. Brbic, and J. Leskovec, “Open-world semi-supervised learning,” in ICLR, 2022

    2022

  44. [52]

    Learning semi-supervised gaussian mixture models for generalized category discovery,

    B. Zhao, X. Wen, and K. Han, “Learning semi-supervised gaussian mixture models for generalized category discovery,” ICCV, 2023

    2023

  45. [53]

    Cdad-net: Bridging domain gaps in generalized category discovery,

    S. B. Rongali, S. Mehrotra, A. Jha, S. Bose, T. Gupta, M. Singha, B. Banerjee, et al. , “Cdad-net: Bridging domain gaps in generalized category discovery,” in CVPR, pp. 2616–2626, 2024

    2024

  46. [54]

    The hungarian method for the assignment problem,

    H. W. Kuhn, “The hungarian method for the assignment problem,” Naval research logistics quarterly , vol. 2, no. 1-2, pp. 83–97, 1955

    1955

Pith tools

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