Pith. sign in

REVIEW 3 major objections 5 minor 49 references

Pro2SAM: Mask Prompt to SAM with Grid Points for Weakly Supervised Object Localization

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a frozen SAM, prompted by a grid of points and steered by a transformer's coarse foreground map, produces fine-grained object localization from image-level labels alone.

desk verdict Pro2SAM's mask-matching idea is promising and the IOU90 gains are striking, but the paper overclaims SOTA and leaves the SAM gallery generation too underspecified to reproduce. read the letter →

arxiv 2505.04905 v1 pith:VZ2ABHKL submitted 2025-05-08 cs.CV

Xi Yang , Songsong Duan , Nannan Wang , Xinbo Gao This is my paper · ORCID
classification cs.CV
keywords weaklysupervisedobjectlocalizationSegmentAnythingModelmaskpromptgridpointGlobalTokenTransformerfine-grainedMaxBoxAccV2classactivationmap
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that a frozen Segment Anything Model (SAM) can supply the fine-grained object boundaries that weakly supervised object localization (WSOL) methods lack. The proposed pipeline, Pro2SAM, first trains a Global Token Transformer (GTFormer) to emit a coarse foreground map from image-level labels alone. It then prompts SAM with a grid of points to create a gallery of candidate masks and selects the mask whose pixel-level overlap with the coarse map is highest. The paper reports that this selection lifts Top-1 localization accuracy to 84.03% on CUB-200-2011 and 66.85% on ILSVRC, with the largest gains at strict IoU thresholds such as IOU90. A reader should care because the result suggests that precise localization can be obtained without any box annotation by pairing a weak class-conditional map with a strong zero-shot segmenter.

What carries the argument

The load-bearing machinery is a three-stage selection pipeline. Stage one is GTFormer, a DeiT-S transformer augmented with four learned global tokens; their attention to patch tokens produces a coarse foreground probability map that is trained with classification, edge, and region losses. Stage two feeds a preset grid of point prompts into SAM's frozen prompt encoder and mask decoder to obtain a mask gallery, where the grid avoids the omissions seen with a single point or box. Stage three computes a pixel-level similarity score between the coarse map and each SAM mask, using logical AND over OR (Eq. 11), and returns the highest-scoring mask as the final localization map. The argument rests on this gallery containing at least one mask that closely covers the true foreground object, so the matching step only has to rank it first.

What would settle it

Take an image set where SAM is known to produce incomplete or noisy masks (for example, translucent objects, heavy occlusion, or out-of-distribution classes), then compare Pro2SAM's IOU90 against a variant that skips SAM and uses the GTFormer map directly. If the two accuracies converge or the GTFormer-only variant wins, the SAM mask gallery lacked a covering mask, falsifying the claim that mask matching is responsible for the gain.

Watch

Extended reading notes

Core claim

The central claim is that SAM's semantic ambiguity—its habit of returning background and distractor masks alongside foreground—can be resolved by a mask-prompt matching scheme. GTFormer's output map serves as a mask prompt; a dense grid of point prompts makes SAM produce many candidate masks; and a logical AND/OR overlap score (Eq. 11) picks the candidate most similar to the prompt. The chosen mask is declared the localization result. On CUB-200-2011 this procedure reaches 48.43% MaxBoxAccV2 at IOU90, and on ILSVRC 45.94%, higher than the compared methods, while also improving Top-1 Loc to 84.03% and 66.85% respectively. The paper claims this shows a frozen SAM, never fine-tuned, can overcome the partial-activation problem of CAM and the boundary-dilation problem of transformer attention maps.

Load-bearing premise

Everything depends on the mask gallery that SAM generates: if no candidate mask accurately covers the full foreground object, no similarity score can select a good localization map, and the reported gains would vanish.

Editorial extensions

If this is right

  • WSOL can be decomposed into a coarse map generator and an off-the-shelf frozen segmenter, so fine-grained localization no longer requires training a model to draw sharp boundaries.
  • Strict-IoU metrics (IOU70, IOU90) become the informative benchmark for WSOL, since the method's advantage over prior work is concentrated at those thresholds.
  • Because SAM is not fine-tuned, the same mask gallery and matching step can be reused with any coarse localization source, including CAMs from CNN classifiers.
  • The method's localization error is upper-bounded by SAM's mask gallery: on any image where SAM fails to segment the object, Pro2SAM cannot recover a better box.
  • The reported IOU90 gains, if they persist, call for re-ranking WSOL methods by MaxBoxAccV2 at high thresholds rather than Top-1 Loc alone.

Reading between the lines

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

  • One could test the coarse map's role by replacing GTFormer with a plain CAM or self-attention map; if the matching step still selects good SAM masks, the transformer's global tokens are not essential to the localization gain.
  • Grid-point density is a promising free parameter: too sparse a grid may miss small objects, and too dense a grid may oversegment, so measuring performance across object scales could expose a scale limit that the current benchmarks do not.
  • The same mask-matching scheme could be applied to unsupervised object discovery, where no image-level labels exist at all, by ranking SAM masks with a self-supervised similarity signal instead of a trained coarse map.
  • The large IOU90 margins suggest a new evaluation regime for WSOL, but they are reported on two datasets; replicating on fine-grained domains beyond birds (e.g., cars, aircraft) would show whether SAM's mask gallery is consistently rich enough.
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

3 major / 5 minor

Summary. The paper proposes Pro2SAM, a weakly supervised object localization pipeline that combines a newly designed Global Token Transformer (GTFormer) with the frozen Segment Anything Model (SAM). GTFormer is trained with image-level labels to produce a coarse foreground map, which is then used as a mask prompt to select, from a set of SAM-generated masks, the mask that best matches the coarse map via a pixel-level intersection/union score. The selected mask is the final localization output. The authors report strong results on CUB-200-2011 and ILSVRC, including 84.03% Top-1 Loc on CUB and 66.85% on ILSVRC, and notably high MaxBoxAccV2 IOU90 scores. The core idea is coherent and the three-stage pipeline is clearly presented, but there are several load-bearing ambiguities and claim inconsistencies that need to be resolved before the paper is publishable.

Significance. If the reported results are reproducible, the paper makes a valuable contribution by showing that a frozen, zero-shot foundation model can be used to refine coarse WSOL activation maps without any additional training. The GTFormer design is a reasonable extension of transformer-based WSOL, and the proposed mask-matching mechanism is simple and effective in principle. The reported IOU90 MaxBoxAccV2 numbers are substantially higher than prior WSOL methods, which is a meaningful advance for fine-grained localization evaluation. However, the significance is currently tempered by (i) an inaccurate state-of-the-art claim in the abstract relative to the paper's own Table 1, and (ii) a central methodological ambiguity in how the SAM mask gallery is generated, without which the experimental results cannot be reproduced or even unambiguously interpreted.

major comments (3)
  1. [§3.3, Eq. (10) and §3.1] Equation (10) defines the mask gallery as M_L_SAM = SAM(I, [point_1, ..., point_G]), implying that all grid points are passed as a single sparse prompt set. In the released SAM interface, a single forward pass through the mask decoder with a set of points returns at most three masks (whole, part, and subpart), not a mask gallery whose size L varies per image. The text's statement that 'SAM uses a complex post-processing operation to exclude some noisy masks' describes the automatic mask generator (amg.py), where grid points are processed as individual prompts over many forward passes, followed by duplicate removal and NMS. If Eq. (10) is taken literally, the gallery collapses to at most three masks per image and the reported IOU90 results cannot be produced; if the automatic mask generator was used, this is the load-bearing implementation detail and it must be specified exactly. Please clarify how the grid prompts are issued, how many forward passes are made per image, and what post-processing parameters (points_per_side, prediction IoU threshold, stability score threshold, NMS settings, crop overlap) are used.
  2. [Abstract and Table 1] The abstract claims state-of-the-art performance on both CUB-200-2011 and ILSVRC, but Table 1 shows that LocLoc reports 84.40% Top-1 Loc and 98.10% GT-Known on CUB, while Pro2SAM reports 84.03% and 95.67%. Pro2SAM is therefore not state-of-the-art on CUB; the claim should be restricted to ILSVRC or stated as second-best on CUB, and the comparison with LocLoc should be discussed in the text rather than only appearing in the table.
  3. [§4.1, Implementation Details] Several parameters essential for reproducibility are missing. The grid-point prompt density G defined in Eq. (10) is never assigned a value or described in terms of spatial density; the SAM checkpoint is only described as 'ViT-H model' without the specific released version; and no details are given for the automatic mask generator settings if that is what was used. In addition, no error bars or number of runs are reported, which is particularly important because Table 1 contains comparisons with differences under one point (e.g., Ours-GTFormer 80.48 vs SAT 80.43 Top-1 Loc on CUB). Please provide the missing settings and report variance or at least multiple-seed results for the main claims.
minor comments (5)
  1. [§4.3, captions of Tables 4 and 5] The table labelled 'T able 4: Ablation studies for numbers of Global Token on ILSVRC dataset' actually contains the (mu, lambda) hyperparameter sweep, while the subsequent Table 5 reports the number of global tokens; the Table 4 caption appears to be a copy-paste error and should be corrected.
  2. [§4.2, paragraph on Fine-grained Localization] The sentence claiming IOU70 and IOU90 improvements of '14.92% and 18.80% compared with the second method TokenCut on CUB' does not match the table: for IOU90, Pro2SAM's 48.43 vs TokenCut's 25.96 is a relative improvement of about 86.5%, not 18.80%. The text should state whether improvements are absolute percentage points or relative percentages and correct the numbers accordingly.
  3. [§4.4, Discussions] The paper promises 'detailed quantitative and qualitative comparisons of single point, single box, and grid points in Supplementary Material,' but no supplementary material is available with this arXiv version. Please either include the supplementary material referenced by the paper or remove the reference.
  4. [Eq. (11)] The summation indices H and W in Eq. (11) are typed in the same style as image height and width, which is confusing; the limits h/p and w/p suggest these are patch indices, so a notation such as i and j or a clarification of the resize/resampling step would improve readability.
  5. [§1, Introduction] The phrase 'we initiatively leverage the capability' is awkward and should be rephrased to 'we are the first to leverage' or 'we propose to leverage' for clarity.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the selected SAM mask is chosen by pixel overlap with a coarse map trained on image-level labels, and all reported localization scores are measured against held-out GT boxes.

full rationale

Pro2SAM's derivation chain is self-contained with respect to the target metric. GTFormer is trained with image-level labels plus the edge/region losses (Eq. 9); the coarse map M_b is not derived from bounding-box annotations. At inference, SAM generates a mask gallery (Eq. 10) and the mask maximizing the pixel-level similarity in Eq. (11) with M_b is selected as the localization map M_Loc. Evaluation metrics (Top-1/Top-5 Loc, MaxBoxAccV2, GT-Known) use held-out GT boxes, which never enter training or mask selection. The final localization is therefore not defined in terms of the evaluation target, and the quality ceiling is explicitly that the selected mask is one of SAM's gallery masks. No load-bearing self-citation appears: reference [12] (a medical-SAM study with overlapping authorship) is a contextual application citation and is not used to justify the mask-matching mechanism. The genuine caveats are reproducibility/specification issues, not circularity: Eq. (10) as written (SAM(I,[point_1,...,point_G])) does not match SAM's released prompt interface, where a single point-set prompt returns at most three masks rather than a mask gallery; the paper mentions complex post-processing and promises a detailed single-point/box/grid comparison in the supplementary material (Section 4.4), but that supplement is not present in this arXiv v1 and the automatic-mask-generator settings are not reported. These omissions weaken reproducibility without making the claimed result equivalent to its inputs.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The central claim rests on three fitted configuration choices (global-token count, loss weights, grid prompt density), four domain assumptions about SAM and the coarse map, and one invented architectural token type. The heaviest load is carried by the assumption that frozen SAM's zero-shot segmentation produces a near-perfect foreground mask for every target image.

free parameters (3)
  • Number of global tokens = 4
    Ablation in Table 5 on ILSVRC shows best Top-1 Loc at 4 tokens (60.2) versus 1 (59.4) and 16 (41.3); this is a hand-chosen architectural constant that directly controls the coarse mask used for matching.
  • Loss weights mu and lambda = mu=1.0, lambda=0.5
    Selected by ablation on ILSVRC (Table 4) and then applied to CUB; these weights control how much edge and region losses shape the coarse map.
  • Grid point prompt density G = not specified
    The paper says preset grid points are fed to SAM but never gives the number or spacing of points; this parameter determines the mask gallery and is likely tuned in the omitted supplementary material.
assumptions (4)
  • domain assumption SAM generates at least one mask closely matching the true foreground object when prompted with grid points.
    The final localization is always one mask from SAM's gallery (Section 3.3, Eq. 10). If SAM oversegments, undersegments, or misses the object, the pipeline cannot recover.
  • domain assumption The GTFormer coarse map is a reliable enough foreground surrogate to select the correct SAM mask.
    Mask matching scores every SAM mask by overlap with the coarse map (Section 3.3, Eq. 11), so a coarse map biased toward background or a discriminative part will select a wrong mask.
  • domain assumption ImageNet-pretrained Deit-S features and image-level class labels are sufficient to train a useful coarse localizer.
    The standard WSOL premise; GTFormer is trained only with class labels and pretrained weights, per Section 4.1 and Eq. (9).
  • domain assumption SAM's zero-shot segmentation transfers to CUB-200-2011 and ILSVRC object classes.
    SAM is frozen and not fine-tuned (Section 4.1); its masks on bird and object photos must align with semantic instances rather than parts or background regions.
invented entities (1)
  • Novel global tokens in GTFormer
    purpose: Four learned tokens that pool fixed-window positions from the downsampled image and are used to query patch tokens, producing the coarse localization map (Eq. 1, Eq. 6).
    Internal architectural device with no independent external handle; its utility is only measured through final benchmark scores.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Pro2SAM: Mask Prompt to SAM with Grid Points for Weakly Supervised Object Localization." pith.science (2026). https://pith.science/paper/VZ2ABHKL

@misc{pith2026250504905,
  author       = {Pith},
  title        = {Pith review of: Pro2SAM: Mask Prompt to SAM with Grid Points for Weakly Supervised Object Localization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VZ2ABHKL}},
  note         = {Machine review of arXiv:2505.04905}
}
read the original abstract

Weakly Supervised Object Localization (WSOL), which aims to localize objects by only using image-level labels, has attracted much attention because of its low annotation cost in real applications. Current studies focus on the Class Activation Map (CAM) of CNN and the self-attention map of transformer to identify the region of objects. However, both CAM and self-attention maps can not learn pixel-level fine-grained information on the foreground objects, which hinders the further advance of WSOL. To address this problem, we initiatively leverage the capability of zero-shot generalization and fine-grained segmentation in Segment Anything Model (SAM) to boost the activation of integral object regions. Further, to alleviate the semantic ambiguity issue accrued in single point prompt-based SAM, we propose an innovative mask prompt to SAM (Pro2SAM) network with grid points for WSOL task. First, we devise a Global Token Transformer (GTFormer) to generate a coarse-grained foreground map as a flexible mask prompt, where the GTFormer jointly embeds patch tokens and novel global tokens to learn foreground semantics. Secondly, we deliver grid points as dense prompts into SAM to maximize the probability of foreground mask, which avoids the lack of objects caused by a single point/box prompt. Finally, we propose a pixel-level similarity metric to come true the mask matching from mask prompt to SAM, where the mask with the highest score is viewed as the final localization map. Experiments show that the proposed Pro2SAM achieves state-of-the-art performance on both CUB-200-2011 and ILSVRC, with 84.03\% and 66.85\% Top-1 Loc, respectively.

Figures

Figures reproduced from arXiv: 2505.04905 by the authors.

Figure 1
Figure 1. (a) Comparison of Ours, CAM-based (CAM [48] and BagCAM [49]), and transformer-based methods (SAT [34] and UOLwRPS [27]). The red and green boxes are GT bounding boxes and predicted bounding boxes, and box prompt, respectively. (b) Ambiguity of SAM, blue bounding box is box prompt and red star is point prompt. hinders the development of computer vision methods in real-world scenarios. To relieve this issue, weakly su… view at source ↗
Figure 2
Figure 2. Overview of our proposed Pro2SAM framework, which comprises three stages: Stage 1) Mask Prompt Generation stage, we propose a Global Token Transformer (GTFormer) to predict coarse foreground maps via only image-level labels; Stage 2) SAM Assist Stage, we input the grid points of the image into SAM as prompts to generate all masks of input images; Stage 3) Mask Matching Stage, we evaluate the similarity scores betwee… view at source ↗
Figure 3
Figure 3. Overview of our proposed GTFormer, which inherits the architecture of trans￾former blocks for ViT [9,29] and adds novel global tokens and global-aware transformer blocks. Furthermore, we propose a global embedding layer to learn global tokens. Global Embedding. As shown in [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Visualization comparison. The ground-truth bounding boxes are in red, and the predicted bounding boxes are in green. Method Top-1 Loc GT-Known SAT [34] 59.9 73.0 Self-Att. Map 43.0 54.7 w/o-GTA 51.1 63.3 GTFormer 60.2 73.2 [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Accurate localization comparison of Pr2SAM and other SOTA methods with diverse IOU thresholds. 4.4 Discussions SAM vs Transformer-based WSOL methods. As we know, transformer￾based WSOL methods usually outperform CAM-based methods via global long￾range dependencies. How…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 42 canonical work pages

  1. [1]

    Achiam, O., Adler, S., Agarwal, S.: Gpt-4 technical report (2023)

    work page 2023

  2. [2]

    In: Proceedings of the ECCV (2022)

    Bai, H., Zhang, R., Wang, J., Wan, X.: Weakly supervised object localization via transformer with implicit spatial calibration. In: Proceedings of the ECCV (2022)

    work page 2022

  3. [3]

    Cao, X., Zheng, X., Shen, Y., Li, K., Chen, J., Lu, Y., Tian, Y.: Locloc: Low-level cuesandlocal-areaguidesforweaklysupervisedobjectlocalization.In:Proceedings of the ACM MM. pp. 5655–5664 (2023)

    work page 2023

  4. [4]

    In: Proceedings of the NeurIPS

    Cen, J., Zhou, Z., Fang, J., Shen, W., Xie, L., Jiang, D., Zhang, X., Tian, Q., et al.: Segment anything in 3d with nerfs. In: Proceedings of the NeurIPS. vol. 36 (2024)

    work page 2024

  5. [5]

    In: Proceedings of the ICCV

    Chen, Z., Ding, J., Cao, L., Shen, Y., Zhang, S., Jiang, G., Ji, R.: Category-aware allocation transformer for weakly supervised object localization. In: Proceedings of the ICCV. pp. 6643–6652 (2023)

    work page 2023

  6. [6]

    In: Proceedings of the AAAI

    Chen, Z., Wang, C., Wang, Y., Jiang, G., Shen, Y., Tai, Y., Wang, C., Zhang, W., Cao, L.: Lctr: On awakening the local continuity of transformer for weakly supervised object localization. In: Proceedings of the AAAI. p. 410–418 (2022)

    work page 2022

  7. [7]

    In: Proceedings of the CVPR

    Choe, J., Oh, S.J., Lee, S., Chun, S., Akata, Z., Shim, H.: Evaluating weakly supervised object localization methods right. In: Proceedings of the CVPR. pp. 3133–3142 (2020)

    work page 2020

  8. [8]

    In: Processing of the CVPR (2019)

    Choe, J., Shim, H.: Attention-based dropout layer for weakly supervised object localization. In: Processing of the CVPR (2019)

    work page 2019

Show all 49 references
  1. [9]

    In: Proceedings of the ICLR (2021)

    Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Houlsby, N.: An image is worth 16x16 words: Transformers for image recognition at scale. In: Proceedings of the ICLR (2021)

    2021

  2. [10]

    arXiv preprint arXiv:2401.09826 (2024)

    Feng, C.B., Lai, Q., Liu, K., Su, H., Vong, C.M.: Boosting few-shot semantic seg- mentation via segment anything model. arXiv preprint arXiv:2401.09826 (2024)

    2024 arXiv

  3. [11]

    In: Proceedings of the CVPR

    Gupta, S., Lakhotia, S., Rawat, A., Tallamraju, R.: Vitol: Vision transformer for weakly supervised object localization. In: Proceedings of the CVPR. pp. 4101–4110 (2022)

    2022

  4. [12]

    Huang, Y., Yang, X., Liu, L., Zhou, H., Chang, A., Zhou, X., Chen, R., Yu, J., Chen, J., Chen, C., Liu, S., Chi, H., Hu, X., Yue, K., Li, L., Grau, V., Fan, D.P., Dong, F., Ni, D.: Segment anything model for medical images? Medical Image Analysis92, 103061 (2024)

    2024

  5. [13]

    In: Proceedings of the NeurIPS

    Ke, L., Ye, M., Danelljan, M., Tai, Y.W., Tang, C.K., Yu, F., et al.: Segment anything in high quality. In: Proceedings of the NeurIPS. vol. 36 (2024) 16 Xi Yang, Songsong Duan, et al

    2024

  6. [14]

    In: Proceedings of NAACL-HLT

    Kenton, J.D.M.W.C., Toutanova, L.K.: Bert: Pre-training of deep bidirectional transformers for language understanding. In: Proceedings of NAACL-HLT. vol. 1, p. 2 (2019)

    2019

  7. [15]

    In: Proceedings of the ICCV

    Kirillov, A., Mintun, E., Ravi, N., Mao, H., Rolland, C., Gustafson, L., Xiao, T., Whitehead, S., Berg, A.C., Lo, W.Y., Dollár, P., Girshick, R.: Segment anything. In: Proceedings of the ICCV. pp. 3992–4003 (2023)

    2023

  8. [16]

    In: Proceedings of the ACL (2020)

    Lewis, M., Liu, Y., Goyal, N., Ghazvininejad, M., Mohamed, A., Levy, O., Stoy- anov, V., Zettlemoyer, L.: Bart: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. In: Proceedings of the ACL (2020)

    2020

  9. [17]

    Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., Guo, B.: Swin transformer:Hierarchical visiontransformerusingshiftedwindows.In:Proceedings of the ICCV. pp. 10012–10022 (2021)

    2021

  10. [18]

    In: Proceedings of the ECCV

    Lu, W., Jia, X., Xie, W., Shen, L., Zhou, Y., Duan, J.: Geometry constrained weakly supervised object localization. In: Proceedings of the ECCV. pp. 481–496 (2020)

    2020

  11. [19]

    In: Processing of the CVPR (2020)

    Mai, J., Yang, M., Luo, W.: Erasing integrated learning: A simple yet effective approach for weakly supervised object localization. In: Processing of the CVPR (2020)

    2020

  12. [20]

    Medical Image Analysis89, 102918 (2023)

    Mazurowski, M.A., Dong, H., Gu, H., Yang, J., Konz, N., Zhang, Y.: Segment anything model for medical image analysis: an experimental study. Medical Image Analysis89, 102918 (2023)

    2023

  13. [21]

    In: Processing of the ICCV (2021)

    Meng, M., Zhang, T., Tian, Q., Zhang, Y., Wu, F.: Foreground activation maps for weakly supervised object localization. In: Processing of the ICCV (2021)

    2021

  14. [22]

    In: Processing of the CVPR (2021)

    Pan, X., Gao, Y., Lin, Z., Tang, F., Dong, W., Yuan, H., Huang, F., Xu, C.: Unveil- ing the potential of structure preserving for weakly supervised object localization. In: Processing of the CVPR (2021)

    2021

  15. [23]

    In: Proceedings of the AAAI

    Pan, Y., Yao, Y., Cao, Y., Chen, C., Lu, X.: Coarse2fine: local consistency aware re- prediction for weakly supervised object localization. In: Proceedings of the AAAI. vol. 37, pp. 2002–2010 (2023)

    2023

  16. [24]

    In: Proceedings of the ICML

    Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., Krueger, G., Sutskever, I.: Learning transferable visual models from natural language supervision. In: Proceedings of the ICML. vol. 139, pp. 8748–8763 (2021)

    2021

  17. [25]

    arXiv preprint arXiv:2307.01197 (2023)

    Rajič, F., Ke, L., Tai, Y.W., Tang, C.K., Danelljan, M., Yu, F.: Segment anything meets point tracking. arXiv preprint arXiv:2307.01197 (2023)

    2023 arXiv

  18. [26]

    International Journal of Computer Vision115, 211–252 (2015)

    Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al.: Imagenet large scale visual recog- nition challenge. International Journal of Computer Vision115, 211–252 (2015)

    2015

  19. [27]

    In: Proceedings of the ICCV

    Song, Y., Jang, S., Katabi, D., Son, J.: Unsupervised object localization with rep- resenter point selection. In: Proceedings of the ICCV. pp. 6534–6544 (2023)

    2023

  20. [28]

    In: Proceedings of the NeurIPS

    T, B., B, M., N, R., et al: Language models are few-shot learners. In: Proceedings of the NeurIPS. pp. 1877–1901 (2020)

    2020

  21. [29]

    In: Proceedings of the ICML

    Touvron, H., Cord, M., Douze, M., Massa, F., Sablayrolles, A., Jégou, H.: Training data-efficient image transformers & distillation through attention. In: Proceedings of the ICML. pp. 10347–10357. PMLR (2021)

    2021

  22. [30]

    In: Proceedings of the CVPR

    Wang, Y., Shen, X., Hu, S.X., Yuan, Y., Crowley, J.L., Vaufreydaz, D.: Self- supervised transformers for unsupervised object discovery using normalized cut. In: Proceedings of the CVPR. pp. 14543–14553 (2022) Pro2SAM 17

    2022

  23. [31]

    In: Proceedings of the CVPR

    Wei, J., Wang, Q., Li, Z., Wang, S., Zhou, S.K., Cui, S.: Shallow feature matters for weakly supervised object localization. In: Proceedings of the CVPR. pp. 5993–6001 (2021)

    2021

  24. [32]

    Welinder, P., Branson, S., Mita, T., Wah, C., Schroff, F., Belongie, S., Perona, P.: Caltech-ucsd birds 200 (2010)

    2010

  25. [33]

    In: Proceedings of the CVPR

    Wu, P., Zhai, W., Cao, Y.: Background activation suppression for weakly super- vised object localization. In: Proceedings of the CVPR. pp. 14228–14237 (2022)

    2022

  26. [34]

    In: Proceedings of the ICCV

    Wu, P., Zhai, W., Cao, Y., Luo, J., Zha, Z.J.: Spatial-aware token for weakly supervised object localization. In: Proceedings of the ICCV. pp. 1844–1854 (2023)

    2023

  27. [35]

    In: Proceedings of the ICCV

    Xie, J., Luo, C., Zhu, X., Jin, Z., Lu, W., Shen, L.: Online refinement of low- level feature based activation map for weakly supervised object localization. In: Proceedings of the ICCV. pp. 132–141 (2021)

    2021

  28. [36]

    In: Proceedings of the CVPR

    Xie, J., Xiang, J., Chen, J., Hou, X., Zhao, X., Shen, L.: C2am: Contrastive learn- ing of class-agnostic activation map for weakly supervised object localization and semantic segmentation. In: Proceedings of the CVPR. pp. 989–998 (2022)

    2022

  29. [37]

    In: Proceedings of the CVPR

    Xu, J., Hou, J., Zhang, Y., Feng, R., Zhao, R.W., Zhang, T., Lu, X., Gao, S.: Cream: Weakly supervised object localization via class re-activation mapping. In: Proceedings of the CVPR. pp. 9437–9446 (2022)

    2022

  30. [38]

    In: Proceed- ings of the CVPR

    Xu, L., Ouyang, W., Bennamoun, M., Boussaid, F., Xu, D.: Learning multi-modal class-specific tokens for weakly supervised dense object localization. In: Proceed- ings of the CVPR. pp. 19596–19605 (2023)

    2023

  31. [39]

    arXiv preprint arXiv:2311.17707 (2023)

    Xu, M., Yin, X., Qiu, L., Liu, Y., Tong, X., Han, X.: Sampro3d: Locating sam prompts in 3d for zero-shot scene segmentation. arXiv preprint arXiv:2311.17707 (2023)

    2023 arXiv

  32. [40]

    In: Proceedings of the ICCV

    Xue, H., Liu, C., Wan, F., Jiao, J., Ji, X., Ye, Q.: Danet: Divergent activation for weakly supervised object localization. In: Proceedings of the ICCV. pp. 6589–6598 (2019)

    2019

  33. [41]

    IEEE Transactions on Geoscience and Remote Sensing61, 1–16 (2023)

    Yan, Z., Li, J., Li, X., Zhou, R., Zhang, W., Feng, Y., Diao, W., Fu, K., Sun, X.: Ringmo-sam: A foundation model for segment anything in multimodal remote- sensing images. IEEE Transactions on Geoscience and Remote Sensing61, 1–16 (2023)

    2023

  34. [42]

    arXiv preprint arXiv:2304.11968 (2023)

    Yang, J., Gao, M., Li, Z., Gao, S., Wang, F., Zheng, F.: Track anything: Segment anything meets videos. arXiv preprint arXiv:2304.11968 (2023)

    2023 arXiv

  35. [43]

    IEEE Transactions on Neural Networks and Learning Systems p

    Yao, Y., Wan, F., Gao, W., Pan, X., Peng, Z., Tian, Q., Ye, Q.: Ts-cam: Token semantic coupled attention map for weakly supervised object localization. IEEE Transactions on Neural Networks and Learning Systems p. 1–13 (Jan 2022)

    2022

  36. [44]

    arXiv preprint arXiv:2304.06790 (2023)

    Yu, T., Feng, R., Feng, R., Liu, J., Jin, X., Zeng, W., Chen, Z.: Inpaint anything: Segment anything meets image inpainting. arXiv preprint arXiv:2304.06790 (2023)

    2023 arXiv

  37. [45]

    In: Proceedings of the CVPR

    Zhang, C.L., Cao, Y.H., Wu, J.: Rethinking the route towards weakly supervised object localization. In: Proceedings of the CVPR. pp. 13460–13469 (2020)

    2020

  38. [46]

    IEEE Transactions on Pattern Analysis and Machine Intelligence44(9), 5866–5885 (2021)

    Zhang, D., Han, J., Cheng, G., Yang, M.H.: Weakly supervised object localization and detection: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence44(9), 5866–5885 (2021)

    2021

  39. [47]

    In: Proceeding of the AAAI (2024)

    Zhang, X., Liu, Y., Lin, Y., Liao, Q., Li, Y.: Uv-sam: Adapting segment anything model for urban village identification. In: Proceeding of the AAAI (2024)

    2024

  40. [48]

    In: Processing of the CVPR (2016)

    Zhou, B., Khosla, A., Lapedriza, A., Oliva, A., Torralba, A.: Learning deep features for discriminative localization. In: Processing of the CVPR (2016)

    2016

  41. [49]

    In: Proceedings of the ECCV

    Zhu,L.,Chen,Q.,Jin,L.,You,Y.,Lu,Y.:Baggingregionalclassificationactivation maps for weakly supervised object localization. In: Proceedings of the ECCV. pp. 176–192 (2022)

    2022

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.