Pith. sign in

REVIEW 5 major objections 5 minor 63 references

No Annotations for Object Detection in Art through Stable Diffusion

T0 review · 5 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Stable Diffusion finds objects in paintings with zero annotations, using cross-attention maps as a detector.

desk verdict Useful first zero-shot results and an honest oracle, but the class-selectivity of the diffusion detector is unproven; needs a wrong-class control. read the letter →

arxiv 2412.06286 v2 pith:H5XSN747 submitted 2024-12-09 cs.CV

Patrick Ramos , Nicolas Gonthier , Selina Khan , Yuta Nakashima , Noa Garcia This is my paper · ORCID
classification cs.CV
keywords objectdetectioninartzero-shotweakly-superviseddiffusionmodelscross-attentionmapsStablenull-textinversionclassproposer
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 introduces NADA, a pipeline for object detection in paintings that needs no bounding-box annotations and, in its zero-shot variant, no labels at all. It claims that the cross-attention maps of Stable Diffusion, obtained by inverting a painting and reconstructing it with a prompt containing a class name, highlight where that class appears, and that thresholding and segmenting those maps yields usable bounding boxes. Combined with a class proposer—a CLIP-based MLP in the weakly-supervised setting or a vision-language model in the zero-shot setting—NADA reports top weakly-supervised detection on ArtDL 2.0 (45.8 AP50) and the first zero-shot art-detection results, beating CLIP- and InstructBLIP-based class proposals with the same detector. The importance is that most art and historical image collections lack bounding-box annotations, so a pipeline that skips them could let digital humanities researchers localize iconographic objects at scale.

What carries the argument

The load-bearing object is the class-conditioned cross-attention map of Stable Diffusion, extracted through null-text inversion and reconstruction. Specifically, the painting is inverted to noise under a prompt $p$ containing the class label $l$, then reconstructed; the cross-attention maps $\{A'_{jk}\}_{jk}$ between each token and image patch are averaged over all $J$ time steps and $K$ attention blocks (Eq. 1), and the maps for the label's tokens are averaged again (Eq. 2) to give $A_l$. A clamp to $[0,1]$, Otsu binarization, and watershed segmentation convert $A_l$ into masks whose bounding boxes are the detections. This machinery transfers the grounding property of diffusion attention from natural images to paintings without fine-tuning any component.

What would settle it

A decisive test would be to run NADA's Oracle detector (perfect class labels) on a set of paintings where each class name is paired with a distractor class of the same visual category—for instance, 'Child Jesus' versus 'Saint John the Baptist as a child' or 'Saint Francis' versus 'Saint Dominic' in the same pose. If the attention maps for the correct class do not concentrate on the intended figure while excluding the distractor, or if the Oracle AP50 with correct labels is not substantially above a random-box baseline, the claim that diffusion attention localizes semantic classes in art would be refuted.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that a frozen Stable Diffusion model can act as a class-conditioned detector for paintings: given a class label, null-text inversion produces noise that, when denoised with a prompt containing that label, yields cross-attention maps in which the label's token attends precisely to the regions of the painting depicting that class. Averaging these maps across layers and time steps (Eq. 1) and across the label's tokens (Eq. 2), then applying Otsu thresholding and watershed segmentation, produces bounding boxes without any training on the target data. The paper reports that this detector, fed by a weakly-supervised class proposer, reaches 45.8 AP50 on ArtDL 2.0, surpassing prior weakly-supervised methods, and that with a zero-shot proposer it reaches 21.8 AP50 on ArtDL 2.0 and 15.1 on IconArt, which the authors state are the first zero-shot object detection results in art. An oracle experiment with perfect class proposals lifts ArtDL 2.0 to 61.3 AP50, showing the detector's localization is sound while the class proposer is the main bottleneck.

Load-bearing premise

The whole detection pipeline depends on the assumption that Stable Diffusion's cross-attention maps highlight the spatial location of the queried class in a painting; this grounding property is inherited from natural images and is not independently verified for art styles or abstract classes.

Editorial extensions

If this is right

  • NADA reaches 45.8 AP50 on ArtDL 2.0 in the weakly-supervised setting, outperforming prior trained detectors without training the detector itself, and stays within 2.8 AP50 of the strongest baseline on IconArt.
  • Zero-shot object detection in art is feasible: with a VLM class proposer, NADA achieves the first zero-shot results on both benchmarks, even outperforming some weakly-supervised methods.
  • With perfect class knowledge (Oracle), NADA's detection AP50 rises to 61.3 on ArtDL 2.0 and 18.7 on IconArt, indicating that the diffusion localizer is accurate and that better class proposers would directly improve detection.
  • The same detector, without retraining, localizes uncommon art-specific objects such as dragons, unicorns, banners, and swords across styles from Renaissance to ukiyo-e and surrealism.
  • Prompt construction materially affects localization, with template prompts working better on ArtDL 2.0 and caption prompts working better on IconArt, so the prompt must be chosen per dataset rather than universally.

Reading between the lines

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

  • If the grounding assumption holds across art styles, diffusion attention maps could serve as a free localization signal for weakly-supervised semantic segmentation in art, not just detection.
  • The class-proposer bottleneck suggests that pairing NADA's detector with stronger zero-shot VLMs or with iconography-aware prompts could push zero-shot AP substantially closer to the 61.3 oracle ceiling.
  • The failure patterns (misidentifying Paul as Jerome, localizing the wrong person as an angel) indicate that attention maps reliably find a person-like figure even when the identity is wrong; a crop-level verification step comparing prompt-to-image similarity could filter such false positives.
  • The same inversion-attention recipe may transfer to other label-scarce domains, such as medical or satellite imagery, provided the diffusion model has seen those image types.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. The paper proposes NADA, a no-annotation pipeline for object detection in art paintings. Given an image and a list of candidate classes, a class proposer (weakly supervised MLP on CLIP features, or zero-shot VLM prompting) selects classes, and a class-conditioned detector built from Stable Diffusion localizes them. The detector performs null-text inversion, reconstructs the image from a prompt containing the class label, averages cross-attention maps over layers and time steps (Eqs. 1-2), and applies Otsu thresholding plus watershed segmentation to obtain bounding boxes. The method is evaluated on ArtDL 2.0 and IconArt in weakly supervised and zero-shot settings, reporting AP50 of 45.8 on ArtDL 2.0 and 15.1 on IconArt for the zero-shot variant, with an Oracle upper-bound experiment isolating the detector.

Significance. If the central mechanism is sound, the paper would provide a training-free way to localize art-specific classes that are absent from natural-image detection datasets, which is valuable for digital humanities applications. The paper has concrete strengths: the Oracle experiment (Sec. 4.3, Tab. 11) properly isolates the detector from the class proposer; the qualitative attention maps show that the diffusion signal can align with depicted subjects; code is released; and the weakly supervised variant beats prior work on ArtDL 2.0. At the same time, the zero-shot claims rest on an unverified class-selectivity assumption, the headline numbers come from test-selected prompt variants, and the zero-shot baselines are not model-size matched. These issues are fixable with additional experiments and re-analysis, so the contribution remains promising but needs revision.

major comments (5)
  1. [Sec. 4.2.1 and Tab. 11] The detector's class-selectivity is not established. The paper itself notes that when the class proposer misclassifies Paul as Jerome, NADA 'is still able to localize the subject' (Sec. 4.2.1), which suggests the attention map tracks a salient person rather than the queried class. This is corroborated by the per-class Oracle results in Tab. 11: person-like classes are much higher (Saint Sebastian 21.0, Mary 20.3) than abstract/symbolic classes (angel 0.3, nudity 5.4). The authors should run a negative control in which the detector is given a wrong class label for the same image and the confidence/box output is measured; if high-confidence boxes persist under wrong-class prompts, the reported AP50 partly measures 'detecting a salient person' rather than 'detecting the queried class', and the zero-shot claim must be substantially qualified.
  2. [Tabs. 2 and 4, Sec. 4.1-4.2] Headline detection numbers are selected as the best prompt construction method per dataset (Tab. 2) and the best combination of VLM prompt and prompt construction per dataset (Tab. 4). Because this selection is done on the test split, the reported AP50 is an upper envelope over variants and is not a prediction of a single fixed pipeline. The authors should either report both variants for every setting or choose the variant on the validation split and then report the corresponding test number, with the selection rule stated before evaluation.
  3. [Sec. 4.2, Tabs. 3-4] The zero-shot comparison is confounded by model capacity: NADA's ZSCP uses LLaVA-NeXT-34B, while the InstructBLIP baseline uses Vicuna-7B, and the CLIP baseline uses ViT-B/32. Even though the same Stable Diffusion detector is used downstream, the class-proposal quality comparison and the resulting detection AP differences cannot be attributed to the prompting strategy. The authors should add a same-scale baseline (e.g., LLaVA-NeXT-7B or InstructBLIP with a comparable 34B backend) or explicitly state and justify the asymmetry as a design choice rather than an algorithmic advantage.
  4. [Sec. 4, Implementation details; Tab. 11] The label remapping for IconArt changes the semantic content of the detection task: 'Saint Sebastien' is detected as 'person', 'child Jesus' as 'baby', and 'nudity' as 'naked person'. Consequently the per-class AP values in Tab. 11 for these classes do not measure detection of the iconographic class as defined by the dataset; they measure detection of a generic superclass. The paper should report unremapped results or explicitly reframe these classes as superclass detection and discuss how this affects the claim of detecting Christian iconography.
  5. [Abstract and Tab. 2] The abstract states that NADA 'outperforms prior work in weakly-supervised detection', but Tab. 2 shows that on IconArt NADA (AP50 13.8) is below Milani (16.6) and MI-Max-HL (14.5). The claim should be restricted to ArtDL 2.0 or otherwise qualified to match the reported results.
minor comments (5)
  1. [Abstract] There is a typo in 'no annotations for d etection in a rt' in the abstract; the spacing inside 'detection' and 'art' should be fixed.
  2. [Sec. 3.2] In the 'Stable Diffusion reconstruction' paragraph, the text says 'we use Stable Diffusion to generate i', but the method actually reconstructs the input image from the inverted noise; 'reconstruct' would be more precise than 'generate'.
  3. [Tab. 6] The ArtDL 2.0 'multiple' column reports AP50 of 0.03 and 0.02, which is essentially zero and strikingly low given the overall AP50 of 21.8; the authors should explain why NADA completely fails on multi-class ArtDL images, since this seems incompatible with the qualitative example of detecting Mary among five people in Fig. 4.
  4. [Appendix D] There is a typo 'Crucufixion' in the text preceding Tab. 11; it should be 'Crucifixion'.
  5. [Sec. 4.2, baselines] The CLIP baseline uses a cosine similarity threshold of 0.28 taken from LAION-5B filtering, but no justification is given for applying this threshold to art images; the authors should report sensitivity to this choice or justify it empirically.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: NADA is evaluated on held-out benchmarks; its diffusion-attention premise is an external testable assumption, not an input-output identity.

full rationale

The derivation chain is not circular. The class-conditioned detector computes bounding boxes by averaging cross-attention maps over layers and timesteps (Eqs. 1–2, Sec. 3.2), then thresholding with Otsu and applying watershed segmentation; the output boxes are never fed back into the inputs, and no component is fitted to bounding-box ground truth. The weakly-supervised class proposer is trained only on image-level labels (Sec. 3.1), which is the standard weakly-supervised setting rather than a disguised use of detection targets. The zero-shot class proposer is a frozen VLM with no training on target datasets, and its prompt and threshold τ are chosen on the validation split (Sec. 4, Implementation details), not on test boxes. The Oracle ablation (Sec. 4.3) supplies correct classes to the same detector, isolating detector quality rather than defining detection as its own input. The paper's reliance on the property that Stable Diffusion attention maps localize a token is an empirical assumption cited to external work ([56]) and is testable by the reported per-class ablations; it is not a result imported from the authors' own prior publications. Self-citations ([2], [13], [62]) appear only in motivation and application contexts and are not load-bearing. The paper explicitly concedes failure modes — e.g., misclassifying Paul as Jerome while still localizing the subject (Sec. 4.2.1) and label generalization confusing classes (Sec. 3.2, Prompt construction) — which are empirical limitations, not circular definitions. No fitted parameter is renamed as a prediction.

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

The central method relies on pretrained model capacities (CLIP, LLaVA, Stable Diffusion) rather than new physical or mathematical entities. Main hand-tuned inputs are label remappings and the score threshold; no new conserved quantities or mediators are introduced.

free parameters (2)
  • ZSCP score threshold tau = 0.5
    Tuned on the IconArt validation split and used to threshold VLM confidence scores for class proposals (Sec. 4 implementation details).
  • IconArt class label remapping = Saint Sebastien to person; child Jesus to baby; nudity to naked person
    Hand-chosen label edits that make class names more concrete for prompt construction; affects which tokens Stable Diffusion attends to (Sec. 3.2 and Appendix C.2).
assumptions (3)
  • domain assumption Stable Diffusion cross-attention maps localize the prompt token's referent in the image.
    Used in Eqs. 1-2 and bounding box extraction; borrowed from prior work [56] and not re-established for art images in this paper.
  • domain assumption Null-text inversion can reproduce the input image from the constructed prompt with enough fidelity to make attention maps meaningful.
    500-step inversion is a standard method [38], but reconstruction fidelity for diverse art styles is not measured; attention maps are taken from this reconstruction.
  • domain assumption Vision-language model knowledge of Christian iconography transfers to these paintings.
    Both weakly supervised (CLIP plus MLP trained on image labels) and zero-shot (LLaVA) class proposers assume pretrained visual and textual knowledge covers the rare classes in ArtDL and IconArt.

how reviews work

0 comments
Cite this review

Pith. "Pith review of No Annotations for Object Detection in Art through Stable Diffusion." pith.science (2026). https://pith.science/paper/H5XSN747

@misc{pith2026241206286,
  author       = {Pith},
  title        = {Pith review of: No Annotations for Object Detection in Art through Stable Diffusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H5XSN747}},
  note         = {Machine review of arXiv:2412.06286}
}
read the original abstract

Object detection in art is a valuable tool for the digital humanities, as it allows for faster identification of objects in artistic and historical images compared to humans. However, annotating such images poses significant challenges due to the need for specialized domain expertise. We present NADA (no annotations for detection in art), a pipeline that leverages diffusion models' art-related knowledge for object detection in paintings without the need for full bounding box supervision. Our method, which supports both weakly-supervised and zero-shot scenarios and does not require any fine-tuning of its pretrained components, consists of a class proposer based on large vision-language models and a class-conditioned detector based on Stable Diffusion. NADA is evaluated on two artwork datasets, ArtDL 2.0 and IconArt, outperforming prior work in weakly-supervised detection, while being the first work for zero-shot object detection in art. Code is available at https://github.com/patrick-john-ramos/nada

Figures

Figures reproduced from arXiv: 2412.06286 by the authors.

Figure 1
Figure 1. Art object detection in the wild with NADA’s class￾conditioned detector. Christian iconography, an annotator must be able to distin￾guish between St. Francis and St. Dominic. While these classes can be distinguished by their associated symbols as described in online iconography databases such as Icon￾class1 , a deep familiarity with these relationships is still needed to annotate efficiently. As a result, existing o… view at source ↗
Figure 2
Figure 2. NADA consists of predicting classes from a painting with a class proposer and extracting bounding boxes for the predicted classes [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Bounding box extraction from attention maps. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: ArtDL 2.0 and IconArt test images overlaid with NADA [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 6
Figure 6. Figure 6: AP50 results on the ArtDL validation detection set for varying thresholds and when using Otsu’s method. understand artistic interpretations of these classes such as a surrealist rendering of melting skull. These results indicate that NADA is capable of bridging the dom…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 54 canonical work pages

  1. [1]

    Toward cross- domain object detection in artwork images using improved yolov5 and xgboosting

    Tasweer Ahmad and Maximilian Schich. Toward cross- domain object detection in artwork images using improved yolov5 and xgboosting. IET Image Processing, 17(8), 2023. 2

    work page 2023

  2. [2]

    Explain me the painting: Multi-topic knowledgeable art description gen- eration

    Zechen Bai, Yuta Nakashima, and Noa Garcia. Explain me the painting: Multi-topic knowledgeable art description gen- eration. In ICCV, 2021. 1

    work page 2021

  3. [3]

    Label-efficient se- mantic segmentation with diffusion models

    Dmitry Baranchuk, Andrey V oynov, Ivan Rubachev, Valentin Khrulkov, and Artem Babenko. Label-efficient se- mantic segmentation with diffusion models. In ICLR, 2022. 2

    2022

  4. [4]

    A comprehensive survey on object detection in visual art: taxonomy and challenge

    Siwar Bengamra, Olfa Mzoughi, Andr ´e Bigand, and Ezzed- dine Zagrouba. A comprehensive survey on object detection in visual art: taxonomy and challenge. Multimedia Tools and Applications, 83(5), 2024. 2

    work page 2024

  5. [5]

    Weakly supervised deep detection networks

    Hakan Bilen and Andrea Vedaldi. Weakly supervised deep detection networks. In CVPR, 2016. 2

    work page 2016

  6. [6]

    Beyond photo-domain object recognition: Benchmarks for the cross-depiction problem

    Hongping Cai, Qi Wu, and Peter Hall. Beyond photo-domain object recognition: Benchmarks for the cross-depiction problem. In CVPRW, 2015. 1

    work page 2015

  7. [7]

    End-to- end object detection with transformers

    Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, and Sergey Zagoruyko. End-to- end object detection with transformers. In ECCV. Springer,

  8. [8]

    Iconographic image captioning for artworks

    Eva Cetinic. Iconographic image captioning for artworks. In ICPRW. Springer, 2021. 1

    work page 2021

Show all 63 references
  1. [9]

    In search of art

    Elliot J Crowley and Andrew Zisserman. In search of art. In VISART Workshop at ECCV. Springer, 2015. 8

    2015

  2. [10]

    The art of detec- tion

    Elliot J Crowley and Andrew Zisserman. The art of detec- tion. In VISART Workshop at ECCV. Springer, 2016. 2

    2016

  3. [11]

    Instructblip: Towards general- purpose vision-language models with instruction tuning

    Wenliang Dai, Junnan Li, Dongxu Li, Anthony Meng Huat Tiong, Junqi Zhao, Weisheng Wang, Boyang Li, Pascale Fung, and Steven Hoi. Instructblip: Towards general- purpose vision-language models with instruction tuning. In NeurIPS, 2023. 6

    2023

  4. [12]

    Instagen: Enhancing object detection by training on synthetic dataset

    Chengjian Feng, Yujie Zhong, Zequn Jie, Weidi Xie, and Lin Ma. Instagen: Enhancing object detection by training on synthetic dataset. In CVPR, 2024. 2

    2024

  5. [13]

    A dataset and baselines for visual question answering on art

    Noa Garcia, Chentao Ye, Zihua Liu, Qingtao Hu, Mayu Otani, Chenhui Chu, Yuta Nakashima, and Teruko Mita- mura. A dataset and baselines for visual question answering on art. In VISART Workshop at ECCV. Springer, 2020. 1

    2020

  6. [14]

    Detecting people in cubist art.AI Matters, 1(3), 2015

    Shiry Ginosar, Daniel Haas, Timothy Brown, and Jitendra Malik. Detecting people in cubist art.AI Matters, 1(3), 2015. 2

    2015

  7. [15]

    Weakly supervised object detection in artworks

    Nicolas Gonthier, Yann Gousseau, Said Ladjal, and Olivier Bonfait. Weakly supervised object detection in artworks. In VISART Workshop at ECCV, 2018. 1, 2, 4, 5, 6, 7, 11, 12

    2018

  8. [16]

    Multiple instance learning on deep features for weakly supervised ob- ject detection with extreme domain shifts

    Nicolas Gonthier, Sa ¨ıd Ladjal, and Yann Gousseau. Multiple instance learning on deep features for weakly supervised ob- ject detection with extreme domain shifts. CVIU, 214, 2022. 2, 5, 6

    2022

  9. [17]

    Cross-depiction problem: Recognition and synthesis of pho- tographs and artwork

    Peter Hall, Hongping Cai, Qi Wu, and Tadeo Corradi. Cross-depiction problem: Recognition and synthesis of pho- tographs and artwork. Computational Visual Media, 1, 2015. 1

    2015

  10. [18]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In CVPR,

  11. [19]

    Denoising diffu- sion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffu- sion probabilistic models. NeurIPS, 33, 2020. 2

    2020

  12. [20]

    One-shot object detection with co-attention and co-excitation

    Ting-I Hsieh, Yi-Chen Lo, Hwann-Tzong Chen, and Tyng- Luh Liu. One-shot object detection with co-attention and co-excitation. NeurIPS, 32, 2019. 2

    2019

  13. [21]

    Comprehensive attention self-distillation for weakly- supervised object detection

    Zeyi Huang, Yang Zou, BVK Kumar, and Dong Huang. Comprehensive attention self-distillation for weakly- supervised object detection. NeurIPS, 33, 2020. 5, 6

    2020

  14. [22]

    Few-shot object de- tection: Application to medieval musicological studies

    Bekkouch Imad Eddine Ibrahim, Victoria Eyharabide, Val´erie Le Page, and Fr ´ed´eric Billiet. Few-shot object de- tection: Application to medieval musicological studies. J. Imaging, 8(2), 2022. 1

    2022

  15. [23]

    Cross-domain weakly-supervised object de- tection through progressive domain adaptation

    Naoto Inoue, Ryosuke Furuta, Toshihiko Yamasaki, and Kiy- oharu Aizawa. Cross-domain weakly-supervised object de- tection through progressive domain adaptation. In CVPR,

  16. [24]

    Diffusion models for zero-shot open-vocabulary segmentation

    Laurynas Karazija, Iro Laina, Andrea Vedaldi, and Christian Rupprecht. Diffusion models for zero-shot open-vocabulary segmentation. In ECCV, 2025. 2

    2025

  17. [25]

    Context-infused visual grounding for art

    Selina Khan and Nanne van Noord. Context-infused visual grounding for art. In VISART Workshop at ECCV, 2024. 2

    2024

  18. [26]

    Text-image align- ment for diffusion-based perception

    Neehar Kondapaneni, Markus Marks, Manuel Knott, Rog´erio Guimaraes, and Pietro Perona. Text-image align- ment for diffusion-based perception. In CVPR, 2024. 2

    2024

  19. [27]

    Grounded language-image pre-training

    Liunian Harold Li, Pengchuan Zhang, Haotian Zhang, Jian- wei Yang, Chunyuan Li, Yiwu Zhong, Lijuan Wang, Lu Yuan, Lei Zhang, Jenq-Neng Hwang, et al. Grounded language-image pre-training. In CVPR, 2022. 1

    2022

  20. [28]

    Open-vocabulary object segmentation with diffusion models

    Ziyi Li, Qinye Zhou, Xiaoyun Zhang, Ya Zhang, Yanfeng Wang, and Weidi Xie. Open-vocabulary object segmentation with diffusion models. In ICCV, 2023. 1, 2

    2023

  21. [29]

    Microsoft coco: Common objects in context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll´ar, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In ECCV. Springer, 2014. 1

    2014

  22. [30]

    Visual instruction tuning

    Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. Visual instruction tuning. NeurIPS, 36, 2024. 5

    2024

  23. [31]

    Swin transformer: Hierarchical vision transformer using shifted windows

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In ICCV, 2021. 1

    2021

  24. [32]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. ICLR, 2019. 4

    2019

  25. [33]

    Data- efficient image captioning of fine art paintings via virtual- real semantic alignment training

    Yue Lu, Chao Guo, Xingyuan Dai, and Fei-Yue Wang. Data- efficient image captioning of fine art paintings via virtual- real semantic alignment training. Neurocomputing, 490,

  26. [34]

    Diffusionseg: Adapting diffusion towards unsupervised object discovery

    Chaofan Ma, Yuhuan Yang, Chen Ju, Fei Zhang, Jinxiang Liu, Yu Wang, Ya Zhang, and Yanfeng Wang. Diffusionseg: Adapting diffusion towards unsupervised object discovery. arXiv preprint arXiv:2303.09813, 2023. 2, 6

    2023 arXiv

  27. [35]

    9 One-shot object detection in heterogeneous artwork datasets

    Prathmesh Madhu, Anna Meyer, Mathias Zinnen, Lara M¨uhrenberg, Dirk Suckow, Torsten Bendschus, Corinna Reinhardt, Peter Bell, Ute Verstegen, Ronak Kosti, et al. 9 One-shot object detection in heterogeneous artwork datasets. In IPTA. IEEE, 2022. 2

    2022

  28. [36]

    Algorithmic ways of seeing: Using object detection to facilitate art exploration

    Louie Meyer, Johanne Engel Aaen, Anitamalina Regitse Tranberg, Peter Kun, Matthias Freiberger, Sebastian Risi, and Anders Sundnes Løvlie. Algorithmic ways of seeing: Using object detection to facilitate art exploration. In CHI,

  29. [37]

    Proposals generation for weakly supervised ob- ject detection in artwork images

    Federico Milani, Nicol `o Oreste Pinciroli Vago, and Piero Fraternali. Proposals generation for weakly supervised ob- ject detection in artwork images. J. Imaging, 8(8), 2022. 1, 2, 3, 4, 5, 6, 7, 11, 12

    2022

  30. [38]

    Null-text inversion for editing real images using guided diffusion models

    Ron Mokady, Amir Hertz, Kfir Aberman, Yael Pritch, and Daniel Cohen-Or. Null-text inversion for editing real images using guided diffusion models. In CVPR, 2023. 4, 5

    2023

  31. [39]

    Compact watershed and pre- emptive slic: On improving trade-offs of superpixel segmen- tation algorithms

    Peer Neubert and Peter Protzel. Compact watershed and pre- emptive slic: On improving trade-offs of superpixel segmen- tation algorithms. In ICPR. IEEE, 2014. 3, 4

    2014

  32. [40]

    Glide: Towards photorealis- tic image generation and editing with text-guided diffusion models

    Alexander Quinn Nichol, Prafulla Dhariwal, Aditya Ramesh, Pranav Shyam, Pamela Mishkin, Bob Mcgrew, Ilya Sutskever, and Mark Chen. Glide: Towards photorealis- tic image generation and editing with text-guided diffusion models. In ICML. PMLR, 2022. 2

    2022

  33. [41]

    A threshold selection method from gray-level histograms

    Nobuyuki Otsu et al. A threshold selection method from gray-level histograms. Automatica, 11(285-296), 1975. 4

    1975

  34. [42]

    Learn- ing transferable visual models from natural language super- vision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learn- ing transferable visual models from natural language super- vision. In ICLR. PMLR, 2021. 2, 6

    2021

  35. [43]

    Hierarchical text-conditional image genera- tion with clip latents

    Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical text-conditional image genera- tion with clip latents. arXiv preprint arXiv:2204.06125, 1(2),

    arXiv

  36. [44]

    You only look once: Unified, real-time object de- tection

    Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi. You only look once: Unified, real-time object de- tection. In CVPR, 2016. 2

    2016

  37. [45]

    Faster r-cnn: Towards real-time object detection with region proposal networks

    Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r-cnn: Towards real-time object detection with region proposal networks. NeurIPS, 28, 2015. 1, 2

    2015

  38. [46]

    Instance-aware, context-focused, and memory- efficient weakly supervised object detection

    Zhongzheng Ren, Zhiding Yu, Xiaodong Yang, Ming- Yu Liu, Yong Jae Lee, Alexander G Schwing, and Jan Kautz. Instance-aware, context-focused, and memory- efficient weakly supervised object detection. InCVPR, 2020. 2

    2020

  39. [47]

    Deart: Dataset of european art

    Artem Reshetnikov, Maria-Cristina Marinescu, and Joaquim More Lopez. Deart: Dataset of european art. In VISART Workshop at ECCV. Springer, 2022. 1

    2022

  40. [48]

    High-resolution image syn- thesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj¨orn Ommer. High-resolution image syn- thesis with latent diffusion models. In CVPR, 2022. 2

    2022

  41. [49]

    Laion-5b: An open large-scale dataset for training next generation image-text models

    Christoph Schuhmann, Romain Beaumont, Richard Vencu, Cade Gordon, Ross Wightman, Mehdi Cherti, Theo Coombes, Aarush Katta, Clayton Mullis, Mitchell Worts- man, et al. Laion-5b: An open large-scale dataset for training next generation image-text models. NeurIPS, 35, 2022. 2, 6

    2022

  42. [50]

    Object discovery via contrastive learning for weakly supervised object detection

    Jinhwan Seo, Wonho Bae, Danica J Sutherland, Junhyug Noh, and Daijin Kim. Object discovery via contrastive learning for weakly supervised object detection. In ECCV. Springer, 2022. 2

    2022

  43. [51]

    Discover- ing visual patterns in art collections with spatially-consistent feature learning

    Xi Shen, Alexei A Efros, and Mathieu Aubry. Discover- ing visual patterns in art collections with spatially-consistent feature learning. In CVPR, 2019. 1

    2019

  44. [52]

    Uwsod: Toward fully-supervised-level capacity weakly supervised object detection

    Yunhang Shen, Rongrong Ji, Zhiwei Chen, Yongjian Wu, and Feiyue Huang. Uwsod: Toward fully-supervised-level capacity weakly supervised object detection. NeurIPS, 33,

  45. [53]

    Automated basin de- lineation from digital elevation models using mathematical morphology

    Pierre J Soille and Marc M Ansoult. Automated basin de- lineation from digital elevation models using mathematical morphology. Signal Processing, 20(2), 1990. 3, 4

    1990

  46. [54]

    Denois- ing diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denois- ing diffusion implicit models. ICLR, 2021. 2

    2021

  47. [55]

    Pcl: Proposal cluster learning for weakly supervised object detection

    Peng Tang, Xinggang Wang, Song Bai, Wei Shen, Xiang Bai, Wenyu Liu, and Alan Yuille. Pcl: Proposal cluster learning for weakly supervised object detection. IEEE TPAMI, 42(1),

  48. [56]

    What the daam: Interpreting stable dif- fusion using cross attention

    Raphael Tang, Linqing Liu, Akshat Pandey, Zhiying Jiang, Gefei Yang, Karun Kumar, Pontus Stenetorp, Jimmy Lin, and Ferhan T ¨ure. What the daam: Interpreting stable dif- fusion using cross attention. In ACL, 2023. 4

    2023

  49. [57]

    C-mil: Continuation multiple instance learning for weakly supervised object detection

    Fang Wan, Chang Liu, Wei Ke, Xiangyang Ji, Jianbin Jiao, and Qixiang Ye. C-mil: Continuation multiple instance learning for weakly supervised object detection. In CVPR,

  50. [58]

    Diffusion model is secretly a training-free open vocabulary semantic segmenter

    Jinglong Wang, Xiawei Li, Jing Zhang, Qingyuan Xu, Qin Zhou, Qian Yu, Lu Sheng, and Dong Xu. Diffusion model is secretly a training-free open vocabulary semantic segmenter. arXiv preprint arXiv:2309.02773, 2023. 2

    2023 arXiv

  51. [59]

    Detecting people in artwork with cnns

    Nicholas Westlake, Hongping Cai, and Peter Hall. Detecting people in artwork with cnns. In VISART Workshop at ECCV. Springer, 2016. 1, 2

    2016

  52. [60]

    Learning graphs to model visual objects across different depictive styles

    Qi Wu, Hongping Cai, and Peter Hall. Learning graphs to model visual objects across different depictive styles. In ECCV. Springer, 2014. 1

    2014

  53. [61]

    Diffumask: Synthesizing images with pixel-level annotations for semantic segmentation using dif- fusion models

    Weijia Wu, Yuzhong Zhao, Mike Zheng Shou, Hong Zhou, and Chunhua Shen. Diffumask: Synthesizing images with pixel-level annotations for semantic segmentation using dif- fusion models. In ICCV, 2023. 2

    2023

  54. [62]

    Not only generative art: Stable diffusion for content-style disentan- glement in art analysis

    Yankun Wu, Yuta Nakashima, and Noa Garcia. Not only generative art: Stable diffusion for content-style disentan- glement in art analysis. In ICMR, 2023. 2

    2023

  55. [63]

    At- tention as annotation: Generating images and pseudo-masks for weakly supervised semantic segmentation with diffusion

    Ryota Yoshihashi, Yuya Otsuka, Tomohiro Tanaka, et al. At- tention as annotation: Generating images and pseudo-masks for weakly supervised semantic segmentation with diffusion. arXiv preprint arXiv:2309.01369, 2023. 2 10 Appendix A. Datasets An overview of the two art object d...

    2023 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.