StackCLIP: Clustering-Driven Stacked Prompt in Zero-Shot Industrial Anomaly Detection

Jianfeng Qiu; Junfa Li; Ke Xu; Shanshan Wang; Yanning Hou; Yanran Ruan

REVIEW 5 major objections 5 minor 1 cited by

StackCLIP: Clustering-Driven Stacked Prompt in Zero-Shot Industrial Anomaly Detection

T0 review · 5 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 StackCLIP claims that stacking semantically related category names into a single prompt beats precise and abstract prompts for zero-shot industrial anomaly detection.

desk verdict A genuine prompt-stacking idea with strong ablations, but the SOTA claim is contradicted by the paper's own table and the key VisA numbers don't reproduce from the stated protocol. read the letter →

arxiv 2506.23577 v2 pith:CRGB5ETM submitted 2025-06-30 cs.CV

Yanning Hou , Yanran Ruan , Junfa Li , Shanshan Wang , Jianfeng Qiu , Ke Xu This is my paper

classification cs.CV

keywords zero-shotanomalydetectionCLIPindustrialpromptlearningstackedpromptsclusteringfeaturealignmentsegmentation

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 is trying to settle what kind of text prompt should accompany CLIP in zero-shot industrial anomaly detection. Existing methods either name the exact category in the prompt, which makes the model focus on the object and overfit, or drop the category name entirely for a generic "object" prompt, which loses category-specific cues such as wood's natural color variation. The proposed middle path is a stacked prompt that groups semantically nearby categories, for example "a photo of a damaged bottle [glass, plastic, textile]". The paper argues that stacked prompts keep enough category information while staying general, and that clustering-driven grouping plus cluster-specific learned alignment layers and prompt regularization push both segmentation and classification ahead of prior zero-shot CLIP-based methods. If true, the result is a single frozen-CLIP system that adapts to new categories without retraining and with better training stability than precise-prompt approaches.

What carries the argument

The central object is the stacked text prompt, a template that appends several category names after the normality or abnormality state word: `a photo of a [state][cls_1][cls_2]...[cls_n]`. The Clustering-Driven Stacked Prompt (CSP) module is the selection mechanism: it embeds category names with the CLIP text encoder, runs k-means on those embeddings, and picks the number of clusters using the cluster-count-penalized intra-cluster variance score of Eq. (2). The Ensemble Feature Alignment (EFA) module is the second load-bearing mechanism: it trains a separate linear projector per cluster and per selected image-encoder layer, then adaptively averages the resulting anomaly maps with attention weights derived from cosine similarity between the test image's CLS token and each cluster text embedding. The Regulating Prompt Learning (RPL) module adds a mean-squared-error loss between learned prompt embeddings and a fixed stacked-prompt embedding; this is what prevents the classification branch's prompt learning from overfitting.

What would settle it

Train the EFA module on VisA with the stacked-prompt template but assign categories to clusters at random, preserving the k-means cluster sizes, and compare AUPRO on MVTec-AD; if random clusters match or exceed the k-means result, the clustering premise is not carrying the reported gain.

Watch

Extended reading notes

Core claim

The central claim is that the limiting factor in CLIP-based zero-shot industrial anomaly detection is the prompt's specificity, and that replacing one category name with a stack of semantically analogous category names is the operative fix. StackCLIP clusters the training category names by their CLIP text embeddings using k-means, builds one stacked prompt per cluster, trains a small linear projection head per cluster and per selected image-encoder layer (layers 6, 12, 18, and 24 of the frozen CLIP encoder) to align image patches with that cluster's text embedding, and at test time weights the heads by cosine similarity between the test image's CLS token and each cluster's text embedding. The paper reports this reaches an AUPRO of 86.4 on MVTec-AD against 85.1 for the strongest prior staged dual-path method and 81.4 for the object-agnostic prompt-learning method, and that it tops all baselines on VisA. It also reports that the stacked prompt alone raises AUPRO by 42.6 points over the precise prompt when only linear heads are trained, and that the RPL regularization term raises classification AUROC from 83.4 to 91.7. The paper concludes that stacked prompts generalize better than precise or abstract prompts, that cluster-specific ensemble alignment improves segmentation, and that regularizing learned prompts toward a fixed stacked-prompt embedding improves classification.

Load-bearing premise

The method assumes that k-means on CLIP text embeddings of category names yields clusters in which one shared stacked prompt and one shared learned linear projection can represent every member category well enough to separate normal from anomalous patches.

Editorial extensions

If this is right

A single frozen CLIP backbone with a few linear heads can be trained on one industrial dataset and transferred zero-shot to another without per-category retraining.
Stacked prompts reduce training instability and overfitting compared to precise prompts, whose AUPRO can drop sharply within a single training epoch.
Classification improves when learned prompts are regularized toward a fixed stacked-prompt embedding rather than trained freely.
As more training categories and cluster-specific linear heads are added, performance generally improves, suggesting the method scales with data diversity.
On the reported benchmarks the largest gains are in anomaly segmentation, with AUPRO, AP, and F1-max improved substantially over prior CLIP-based zero-shot detectors.

Reading between the lines

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

Since the text encoder is frozen, the clusters inherit CLIP's embedding biases; the same stacked-prompt recipe could plausibly transfer to other vision-language backbones, but the k-means cluster quality would need to be re-measured per backbone.
The large AUPRO jump from precise to stacked prompts with identical linear heads suggests much of the gain may come from the prompt template itself; ablating with random but valid category stacks would isolate how much semantic clustering contributes.
The paper's stated difficulty with ambiguous cluster boundaries implies that soft clustering or learned category-to-cluster assignment could extend the method to datasets whose category names are not cleanly separable in text embedding space.
For industrial deployment, the frozen-backbone design keeps adaptation cheap, but categories that are hard to describe in text would likely need auxiliary image-only or multi-modal descriptors, which the paper leaves to future work.

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.

Desk Editor's Note

read the letter

First thing you should know: the core idea—stacking category names into one prompt instead of using precise or abstract prompts—is real and the paper shows it convincingly in the ablations. The training curves in Fig. 4 are clear: stacked prompts beat precise prompts by a wide margin on MVTec-AD when trained on VisA, and also beat abstract prompts by a smaller but consistent margin. That is a useful, cheap improvement for zero-shot industrial inspection. The EFA module, with cluster-specific linear heads and attention-weighted ensembling, is a reasonable extension and the clustering visualization shows it does group semantically related categories.

But the paper's headline claim—state of the art on both detection and segmentation across seven datasets—is not supported by what's reported. Table 1 only compares baselines on MVTec-AD and VisA. The other five datasets appear only in Table 2, which contains StackCLIP's own numbers and no baselines, so calling that SOTA is just an assertion. More troubling, Table 1 itself undercuts the claim: on MVTec-AD classification, StackCLIP gets 91.7 AUROC and 92.7 F1-max, both below SDP+ (92.2 and 93.4). The text admits this but still says 'state-of-the-art across both segmentation and classification tasks' a few lines earlier. That's an internal contradiction.

The VisA row in Table 1 is also not reproducible from the stated protocol. The paper says when testing on VisA, it uses the MVTec-AD training set and a single linear-layer training stage. But Table 2's closest configurations (MVTec+MPDD, MVTec+DTD, MVTec+DAGM) give different numbers, and no MVTec-only row is reported. The AUPRO of 89.8 and AP of 28.0 in Table 1 don't match any of those rows. So the headline VisA result cannot be reconstructed from the paper's own tables.

The cluster-count selection equation (Eq. 2) is never used to justify the n* chosen for Table 1. Instead the experiments just try cluster numbers 1, 2, and 3 and report the best. That's post-hoc selection, and Table 2 shows performance varies with cluster count by a few points, so this matters. No error bars anywhere, which is common in this literature but worth noting given the small differences (1–3 AUPRO points) that the SOTA claims rest on. The references also have some sloppiness (duplicate entries, an 'et al.' in an author slot), but that's minor.

Bottom line: this is a solid, incremental idea that needs a serious referee and a major cleanup before publication. The stacked-prompt comparison is worth publishing; the SOTA claim needs to be scaled back to 'competitive on the two datasets we compared' and the VisA numbers need to be reconciled with the stated protocol. I'd send it to review, not desk-reject, but I'd make the authors fix the reproducibility issues before I'd trust the headline numbers.

Referee Report

5 major / 5 minor

Summary. StackCLIP proposes a zero-shot industrial anomaly detection method built on 'stacked prompts,' in which multiple category names are concatenated into a single text prompt rather than using precise per-category prompts or fully abstract 'object' prompts. The method has three trainable components on top of a frozen CLIP backbone: Clustering-Driven Stacked Prompts (CSP), which groups category names by k-means on CLIP text embeddings; Ensemble Feature Alignment (EFA), which learns cluster-specific linear projection heads on multi-level image features; and Regulating Prompt Learning (RPL), which regularizes learnable prompts with a stacked-prompt reference embedding. The paper reports experiments on seven industrial datasets and claims state-of-the-art performance in both zero-shot anomaly classification and segmentation, with detailed ablations on prompt type, cluster number, and the RPL module.

Significance. The stacked-prompt idea is a genuinely different middle ground between precise and abstract prompts, and the paper's internal ablation in Section 4.3, where precise, abstract, and stacked prompts are compared under the same linear-layer training protocol, is a clean and informative experiment. The framework is also conceptually simple and computationally light, using only two training epochs on a single GPU. If the reported gains are reproducible and the claims are restricted to what the evidence supports, the CSP/EFA/RPL design could be a useful contribution to zero-shot industrial anomaly detection. However, the paper currently overclaims: the state-of-the-art assertion is contradicted on one headline dataset, is not backed by baseline comparisons on five of the seven claimed datasets, and at least one headline number is not reproducible from the stated protocol.

major comments (5)

[§4.2, Table 1] The abstract and Section 4.2 claim state-of-the-art performance in both segmentation and classification, but Table 1 shows that on MVTec-AD the proposed method's image-level AUROC (91.7) and F1-max (92.7) are 0.5 and 0.7 points below SDP+ (92.2 and 93.4), which the text itself acknowledges. The classification SOTA claim is therefore internally contradicted on one of the two headline datasets. The claim should be restricted to segmentation, or additional classification results on the other five datasets should be provided.
[§4.1 Implementation Details, §4.2 Table 1, §4.3 Table 2] The VisA row in Table 1 (AUPRO 89.8, AP 28.0, F1-max 34.2) cannot be reproduced from the stated protocol. Implementation Details says that when testing on VisA, the MVTec-AD training set is used, but Table 2 contains no MVTec-only training row for VisA; the VisA-test rows are trained on MVTec+MPDD, MVTec+DTD, and MVTec+DAGM, and none of them reports (89.8, 28.0, 34.2). The exact training mixture and cluster number used for the Table 1 VisA row must be specified, otherwise the headline VisA SOTA result is not reproducible.
[§4.2, Table 2] The seven-dataset state-of-the-art claim is not supported by the reported evidence. Table 1 compares against baselines only on MVTec-AD and VisA, while Table 2 reports only StackCLIP's own pixel-level numbers for DTD-Synthetic, MPDD, DAGM, BTAD, and SDD. Without baseline comparisons on those five datasets, the abstract's claim of 'state-of-the-art performance' across seven industrial anomaly detection datasets is unsubstantiated. Either add the missing baseline comparisons or narrow the claim.
[§3.2, Eq. (2); §4.3, Table 2] The cluster-number selection mechanism in Eq. (2) is not actually used in the experimental protocol. Table 2 reports results for cluster numbers 1, 2, and 3 and shows that performance varies substantially with the cluster count, yet the paper never states which cluster number Eq. (2) selects for each training mixture, nor whether the numbers used in Table 1 were selected a priori. Without this information, the best-performing cluster counts in Table 2 could be a post hoc selection, and the scoring mechanism in Eq. (2) is not validated. The authors should report the n* chosen by Eq. (2) for each setting and compare it with the grid of cluster numbers.
[§4.2, Tables 1 and 2] No error bars or multiple-seed results are reported anywhere in the paper. Several headline comparisons are close, such as the 0.5 AUROC gap on MVTec-AD classification and the 0.4-1.3 AUPRO differences among Table 2 rows, so without run-to-run variance it is impossible to assess whether the reported improvements are statistically meaningful. Reporting means and standard deviations over at least three seeds is needed for the main tables.

minor comments (5)

[§3.3, Eq. (10)] The focal loss equation uses an undefined symbol M_f in the first term; this appears to be a typo for M^ℓ_{i,j} and should be corrected.
[§3.3, Eq. (14) and surrounding text] The inference description defines the test-cluster-specific text embedding t_{t,i}, but Eq. (14) computes attention weights using t_i from Eq. (8); the paper should clarify which text embedding is actually used for the weights.
[§2.2, Reference [15]] Reference [15] is cited for prompt learning, but the bibliography entry is a knowledge-graph completion paper; the citation appears mismatched and should be replaced with an appropriate prompt-learning reference.
[§4.2, first paragraph] The method name 'APRIL-GAN' is spelled 'APIRL-GAN' in the text; the spelling should be made consistent with Table 1.
[Table 1, header] The column header 'RankTraning' appears to be an accidental merge of 'Rank' and 'Training'; the meaning of the last numeric column should be clarified or split into separate headers.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the method's components are defined by their own equations and validated by disjoint-dataset zero-shot transfer, with self-citations confined to related work.

full rationale

StackCLIP's derivation chain is not circular. The central modules CSP, EFA, and RPL are specified by explicit equations (Eqs. 1-18) and trained with frozen CLIP parameters on disjoint auxiliary datasets (VisA for MVTec-AD tests and MVTec-AD for VisA tests), so the reported segmentation and classification numbers are genuine cross-dataset transfer results rather than quantities fitted to the test set. The cluster-count scoring mechanism (Eq. 2) is a stated selection rule based on intra-cluster variance with a penalty term; it is not defined in terms of the final anomaly-detection metrics. The only self-citations ([14], [15]) appear in related-work enumerations and do not supply any load-bearing premise, uniqueness claim, or ansatz. The claimed state-of-the-art status is an empirical benchmark comparison, not a derivation that reduces to its inputs. Concerns about the unexplained VisA row in Table 1, the classification gap versus SDP+ on MVTec-AD, and the under-specified cluster-number selection for the headline results are matters of verification and reproducibility, not circularity.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

No invented physical entities. The design depends on hand-chosen hyperparameters such as cluster penalty, selected layers, loss weights, and prompt length and depth, and on domain assumptions about CLIP feature geometry and cross-dataset transfer. None of these are machine-checked.

free parameters (5)

cluster-count penalty coefficient = 0.1 in lambda(n) = 0.1 * exp(n)
Chosen ad hoc in Eq. (2); controls how many clusters CSP selects and therefore all downstream cluster-specific linear layers.
cluster number n* = 2 for VisA-based runs; 1/2/3 for merged-dataset runs
Selected via the scoring formula, but Table 2 reports multiple cluster counts and final Table 1 results appear to use a particular count without a described validation protocol.
multi-level feature layers = {6, 12, 18, 24}
Selected by hand for multi-level anomaly maps; all final results depend on these layers.
focal and dice loss hyperparameters = alpha=1, gamma=2, epsilon=1
Set manually in Eqs. (10) and (11); not ablated in the paper.
prompt learning hyperparameters = length 12, depth 9, per-layer length 20
Chosen by hand in Section 4.1; the appendix is said to analyze them but is not present in the provided text.

assumptions (4)

domain assumption CLIP text-encoder embeddings of category names are similar enough that k-means clusters correspond to groups with shared visual anomaly semantics.
CSP in Sec. 3.2 relies on this for clustering; if clusters are semantically inconsistent, shared stacked prompts and linear heads cannot align.
domain assumption Auxiliary dataset training with ground-truth anomaly masks transfers to unseen target categories.
Implementation Details in Sec. 4.1 train on VisA for MVTec testing and vice versa; the entire zero-shot evaluation assumes this transfer.
ad hoc to paper Stacking several category names in a single prompt is processed by the CLIP text encoder as a coherent concept without losing category-specific cues.
Core prompt design assumption in Eq. (1); the paper validates it empirically but does not justify it mechanistically.
domain assumption Standard CLIP ViT-L/14@336px backbone is a sufficient feature extractor for pixel-level anomaly localization.
All features come from frozen CLIP layers 6, 12, 18, and 24; no evidence is provided that these layers preserve enough spatial resolution for segmentation.

how reviews work

0 comments

Cite this review

Pith. "Pith review of StackCLIP: Clustering-Driven Stacked Prompt in Zero-Shot Industrial Anomaly Detection." pith.science (2026). https://pith.science/paper/CRGB5ETM

@misc{pith2026250623577,
  author       = {Pith},
  title        = {Pith review of: StackCLIP: Clustering-Driven Stacked Prompt in Zero-Shot Industrial Anomaly Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CRGB5ETM}},
  note         = {Machine review of arXiv:2506.23577}
}

read the original abstract

Enhancing the alignment between text and image features in the CLIP model is a critical challenge in zero-shot industrial anomaly detection tasks. Recent studies predominantly utilize specific category prompts during pretraining, which can cause overfitting to the training categories and limit model generalization. To address this, we propose a method that transforms category names through multicategory name stacking to create stacked prompts, forming the basis of our StackCLIP model. Our approach introduces two key components. The Clustering-Driven Stacked Prompts (CSP) module constructs generic prompts by stacking semantically analogous categories, while utilizing multi-object textual feature fusion to amplify discriminative anomalies among similar objects. The Ensemble Feature Alignment (EFA) module trains knowledge-specific linear layers tailored for each stack cluster and adaptively integrates them based on the attributes of test categories. These modules work together to deliver superior training speed, stability, and convergence, significantly boosting anomaly segmentation performance. Additionally, our stacked prompt framework offers robust generalization across classification tasks. To further improve performance, we introduce the Regulating Prompt Learning (RPL) module, which leverages the generalization power of stacked prompts to refine prompt learning, elevating results in anomaly detection classification tasks. Extensive testing on seven industrial anomaly detection datasets demonstrates that our method achieves state-of-the-art performance in both zero-shot anomaly detection and segmentation tasks.

Figures

Figures reproduced from arXiv: 2506.23577 by the authors.

**Figure 2.** Overview of StackCLIP To improve image-text feature alignment for anomaly detection, StackCLIP introduces Clustering [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Visual Comparison of Zero-Shot Anomaly Segmentation and Test Data Selection Patterns [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Performance comparison of different prompt strategies across evaluation metrics [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: Clustering results of the VisA and three mixed datasets after the Clustering-Driven Stacked Prompt (CSP) module. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Closed form perturbative relativistic modifications to wave-packet dynamics in the quantum harmonic oscillator
quant-ph 2026-03 unverdicted novelty 5.0 of 10

Closed-form O(1/c²) relativistic corrections to QHO wave-packet widths, variances, and uncertainty products leave minimum-uncertainty saturation intact and become percent-level for 1–10 keV electron confinement.

Reference graph

Works this paper leans on

44 extracted references · 32 canonical work pages · cited by 1 Pith paper

[1]

Toshimichi Aota, Lloyd Teh Tzer Tong, and Takayuki Okatani. 2023. Zero-shot versus many-shot: Unsupervised texture anomaly detection. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision . 5564–5572

work page 2023
[2]

Müller, and Bernhard Kainz

Matthew Baugh, James Batten, Johanna P. Müller, and Bernhard Kainz. 2023. Zero-Shot Anomaly Detection with Pre-trained Segmentation Models. CoRR abs/2306.09269 (2023)

arXiv 2023
[3]

Paul Bergmann, Michael Fauser, David Sattlegger, and Carsten Steger. 2019. MVTec AD — A Comprehensive Real-World Dataset for Unsupervised Anom- aly Detection. In 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

work page 2019
[4]

Yunkang Cao, Qian Wan, Weiming Shen, and Liang Gao. 2022. Informative knowledge distillation for image anomaly segmentation. Knowl. Based Syst. 248 (2022), 108846

work page 2022
[5]

Yunkang Cao, Xiaohao Xu, Chen Sun, Yuqi Cheng, Zongwei Du, Liang Gao, and Weiming Shen. 2023. Segment Any Anomaly without Training via Hybrid Prompt Regularization. CoRR abs/2305.10724 (2023)

arXiv 2023
[6]

Yunkang Cao, Jiangning Zhang, Luca Frittoli, Yuqi Cheng, Weiming Shen, and Giacomo Boracchi. 2024. AdaCLIP: Adapting CLIP with Hybrid Learnable Prompts for Zero-Shot Anomaly Detection

work page 2024
[7]

Xuhai Chen and et al. 2023. A Zero-/Few-Shot Anomaly Classification and Segmentation Method for CVPR 2023 V AND Workshop Challenge Tracks 1&2: 1st Place on Zero-shot AD and 4th Place on Few-shot AD. CoRR abs/2305.17382 (2023)

arXiv 2023
[8]

Xuhai Chen, Jiangning Zhang, Guanzhong Tian, Haoyang He, Wuhao Zhang, Yabiao Wang, Chengjie Wang, Yunsheng Wu, and Yong Liu. 2023. CLIP-AD: A Language-Guided Staged Dual-Path Model for Zero-shot Anomaly Detection. CoRR abs/2311.00453 (2023)

arXiv 2023

Show all 44 references

[10]

Hanqiu Deng and Xingyu Li. 2022. Anomaly Detection via Reverse Distillation from One-Class Embedding. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR. 9727–9736

2022
[11]

Hanqiu Deng, Zhaoxiang Zhang, Jinan Bao, and Xingyu Li. 2023. AnoVL: Adapting Vision-Language Models for Unified Zero-shot Anomaly Localization. CoRR abs/2308.15939 (2023)

2023 arXiv
[12]

Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xi- aohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. 2021. An Image is Worth 16x16 Words: Transformers for Image Recogn...

2021
[13]

Dong Gong, Lingqiao Liu, Vuong Le, Budhaditya Saha, Moussa Reda Mansour, Svetha Venkatesh, and Anton van den Hengel. 2019. Memorizing Normality to Detect Anomaly: Memory-Augmented Deep Autoencoder for Unsupervised Anomaly Detection. In IEEE/CVF International Conference on Comp...

2019
[14]

Yanning Hou, Ke Xu, Junfa Li, Yanran Ruan, and Jianfeng Qiu. 2024. Enhancing Zero-Shot Anomaly Detection: CLIP-SAM Collaboration with Cascaded Prompts. In Chinese Conference on Pattern Recognition and Computer Vision . https: //api.semanticscholar.org/CorpusID:274233331

2024
[15]

Yanning Hou, Sihang Zhou, Ke Liang, Lingyuan Meng, Xiaoshu Chen, Ke Xu, Siwei Wang, Xinwang Liu, and Jian Huang. 2025. Soft Reasoning Paths for Knowledge Graph Completion. arXiv preprint arXiv:2505.03285 (2025)

2025
[16]

Jongheon Jeong, Yang Zou, Taewan Kim, Dongqing Zhang, Avinash Ravichan- dran, and Onkar Dabeer. 2023. WinCLIP: Zero-/Few-Shot Anomaly Classification and Segmentation. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR. IEEE, 19606–19616

2023
[17]

Stepan Jezek, Martin Jonak, Radim Burget, Pavel Dvorak, and Milos Skotak
[18]

Muhammad Uzair Khattak, Syed Talal Wasim, Muzammal Naseer, Salman Khan, Ming-Hsuan Yang, and Fahad Shahbaz Khan. 2023. Self-regulating Prompts: Foundational Model Adaptation without Forgetting. In IEEE/CVF International Conference on Computer Vision, ICCV. 15144–15154

2023
[19]

Muhammad Uzair Khattak, Syed Talal Wasim, Muzammal Naseer, Salman Khan, Ming-Hsuan Yang, and Fahad Shahbaz Khan. 2023. Self-regulating Prompts: Foundational Model Adaptation without Forgetting. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

2023
[20]

Junnan Li, Dongxu Li, Caiming Xiong, and Steven C. H. Hoi. 2022. BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Un- derstanding and Generation. In International Conference on Machine Learning, ICML 2022, 17-23 July 2022, Baltimore, Maryland, USA , ...

2022
[21]

Shengze Li, Jianjian Cao, Peng Ye, Yuhan Ding, Chongjun Tu, and Tao Chen
[22]

Yi Li, Hualiang Wang, Yiqun Duan, and Xiaomeng Li. 2023. CLIP Surgery for Better Explainability with Enhancement in Open-V ocabulary Tasks. CoRR abs/2304.05653 (2023)

2023 arXiv
[23]

Yi Li, Hualiang Wang, Yiqun Duan, and Xiaomeng Li. 2023. CLIP Surgery for Better Explainability with Enhancement in Open-V ocabulary Tasks. CoRR (2023)

2023
[24]

Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, Jie Yang, Chun- yuan Li, Jianwei Yang, Hang Su, Jun Zhu, and Lei Zhang. 2023. Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detec- tion. CoRR (2023)

2023
[25]

Fabio Valerio Massoli, Fabrizio Falchi, Alperen Kantarci, Seymanur Akti, Hazim Kemal Ekenel, and Giuseppe Amato. 2022. MOCCA: Multilayer One- Class Classification for Anomaly Detection. IEEE Trans. Neural Networks Learn. Syst. 33, 6 (2022), 2313–2323

2022
[26]

Pankaj Mishra, Riccardo Verk, Daniele Fornasier, Claudio Piciarelli, and Gian Luca Foresti. 2021. VT-ADL: A Vision Transformer Network for Image Anomaly Detection and Localization. In 30th IEEE International Symposium on Industrial Electronics, ISIE 2021, Kyoto, Japan, June 20...

2021
[27]

Shancong Mou, Xiaoyi Gu, Meng Cao, Haoping Bai, Ping Huang, Jiulong Shan, and Jianjun Shi. 2023. RGI: robust GAN-inversion for mask-free image inpainting and unsupervised pixel-wise anomaly detection. In The Eleventh International Conference on Learning Representations, ICLR 2...

2023
[28]

Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. 2021. Learning Transferable Visual Models From Natural Language Supervision. In Proceedings...

2021
[29]

Yongming Rao, Wenliang Zhao, Guangyi Chen, Yansong Tang, Zheng Zhu, Guan Huang, Jie Zhou, and Jiwen Lu. 2022. DenseCLIP: Language-Guided Dense Prediction with Context-Aware Prompting. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022, New Orleans, LA,...

2022
[30]

Karsten Roth, Latha Pemula, Joaquin Zepeda, Bernhard Schölkopf, Thomas Brox, and Peter V . Gehler. 2022. Towards Total Recall in Industrial Anomaly Detection. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR Conference acronym ’XX, June 03–05, 2018, Wood...

2022
[31]

Kihyuk Sohn, Chun-Liang Li, Jinsung Yoon, Minho Jin, and Tomas Pfister. 2021. Learning and Evaluating Representations for Deep One-Class Classification. In International Conference on Learning Representations, ICLR

2021
[32]

Domen Tabernik, Samo Sela, Jure Skvarc, and Danijel Skocaj. 2020. Segmentation-based deep-learning approach for surface-defect detection. J. Intell. Manuf. 31 (2020)

2020
[33]

Yue Wang, Jinlong Peng, Jiangning Zhang, Ran Yi, Yabiao Wang, and Chengjie Wang. 2023. Multimodal Industrial Anomaly Detection via Hybrid Fusion. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023, V ancouver , BC, Canada, June 17-24, 2023

2023
[34]

Matthias Wieler and Tobias Hahn. 2007. Weakly supervised learning for industrial optical inspection. In DAGM symposium in. 11

2007
[35]

Guoyang Xie, Jinbao Wang, Jiaqi Liu, Yaochu Jin, and Feng Zheng. 2023. Pushing the Limits of Fewshot Anomaly Detection in Industry Vision: Graphcore. In The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5, 2023

2023
[36]

Jihun Yi and Sungroh Yoon. 2020. Patch SVDD: Patch-Level SVDD for Anomaly Detection and Segmentation. In 15th Asian Conference on Computer Vision ACCV (Lecture Notes in Computer Science, V ol. 12627), Hiroshi Ishikawa, Cheng-Lin Liu, Tomás Pajdla, and Jianbo Shi (Eds.). 375–390

2020
[37]

Ni, and Heung-Yeung Shum

Hao Zhang, Feng Li, Shilong Liu, Lei Zhang, Hang Su, Jun Zhu, Lionel M. Ni, and Heung-Yeung Shum. 2023. DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection. In The Eleventh International Conference on Learning Representations, ICLR

2023
[38]

Kaiyang Zhou, Jingkang Yang, Chen Change Loy, and Ziwei Liu. 2021. Learning to Prompt for Vision-Language Models. CoRR abs/2109.01134 (2021)

2021 arXiv
[39]

Qihang Zhou, Guansong Pang, Yu Tian, Shibo He, and Jiming Chen. 2024. Anoma- lyCLIP: Object-agnostic Prompt Learning for Zero-shot Anomaly Detection. In The Twelfth International Conference on Learning Representations (ICLR) . 1–33

2024
[40]

Jiale Zhu, Peiyi Yan, Jielin Jiang, Yan Cui, and Xiaolong Xu. 2024. Asymmetric Teacher-Student Feature Pyramid Matching for Industrial Anomaly Detection. IEEE Trans. Instrum. Meas. 73 (2024), 1–13

2024
[41]

Yang Zou, Jongheon Jeong, Latha Pemula, Dongqing Zhang, and Onkar Dabeer
[45]

In Computer Vision - ECCV 2022 - 17th European Conference, Tel Aviv, V ol

SPot-the-Difference Self-supervised Pre-training for Anomaly Detection and Segmentation. In Computer Vision - ECCV 2022 - 17th European Conference, Tel Aviv, V ol. 13690. 392–408

2022
[2021]

In 13th International Congress on Ultra Modern Telecommunications and Control Systems and Workshops, ICUMT 2021, Brno, Czech Republic, October 25-27, 2021

Deep learning-based defect detection of metal parts: evaluating current methods in complex conditions. In 13th International Congress on Ultra Modern Telecommunications and Control Systems and Workshops, ICUMT 2021, Brno, Czech Republic, October 25-27, 2021 . 66–71

2021
[2022]

https://doi.org/10.1109/CVPR52688.2022.01755

IEEE, 18061–18070. https://doi.org/10.1109/CVPR52688.2022.01755

2022
[2024]

ArXiv abs/2401.12665 (2024)

ClipSAM: CLIP and SAM Collaboration for Zero-Shot Anomaly Segmen- tation. ArXiv abs/2401.12665 (2024). https://api.semanticscholar.org/CorpusID: 267095167

2024 arXiv

Pith tools

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

[1] [1]

Toshimichi Aota, Lloyd Teh Tzer Tong, and Takayuki Okatani. 2023. Zero-shot versus many-shot: Unsupervised texture anomaly detection. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision . 5564–5572

work page 2023

[2] [2]

Müller, and Bernhard Kainz

Matthew Baugh, James Batten, Johanna P. Müller, and Bernhard Kainz. 2023. Zero-Shot Anomaly Detection with Pre-trained Segmentation Models. CoRR abs/2306.09269 (2023)

arXiv 2023

[3] [3]

Paul Bergmann, Michael Fauser, David Sattlegger, and Carsten Steger. 2019. MVTec AD — A Comprehensive Real-World Dataset for Unsupervised Anom- aly Detection. In 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

work page 2019

[4] [4]

Yunkang Cao, Qian Wan, Weiming Shen, and Liang Gao. 2022. Informative knowledge distillation for image anomaly segmentation. Knowl. Based Syst. 248 (2022), 108846

work page 2022

[5] [5]

Yunkang Cao, Xiaohao Xu, Chen Sun, Yuqi Cheng, Zongwei Du, Liang Gao, and Weiming Shen. 2023. Segment Any Anomaly without Training via Hybrid Prompt Regularization. CoRR abs/2305.10724 (2023)

arXiv 2023

[6] [6]

Yunkang Cao, Jiangning Zhang, Luca Frittoli, Yuqi Cheng, Weiming Shen, and Giacomo Boracchi. 2024. AdaCLIP: Adapting CLIP with Hybrid Learnable Prompts for Zero-Shot Anomaly Detection

work page 2024

[7] [7]

Xuhai Chen and et al. 2023. A Zero-/Few-Shot Anomaly Classification and Segmentation Method for CVPR 2023 V AND Workshop Challenge Tracks 1&2: 1st Place on Zero-shot AD and 4th Place on Few-shot AD. CoRR abs/2305.17382 (2023)

arXiv 2023

[8] [8]

Xuhai Chen, Jiangning Zhang, Guanzhong Tian, Haoyang He, Wuhao Zhang, Yabiao Wang, Chengjie Wang, Yunsheng Wu, and Yong Liu. 2023. CLIP-AD: A Language-Guided Staged Dual-Path Model for Zero-shot Anomaly Detection. CoRR abs/2311.00453 (2023)

arXiv 2023