Pith. sign in

REVIEW 3 major objections 5 minor 27 references

Q-Former Autoencoder: A Modern Framework for Medical Anomaly Detection

T0 review · 3 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 QFAE shows that an autoencoder built from a frozen vision foundation model, a Q-Former bottleneck, and a masked-autoencoder perceptual loss achieves top AUROC on three medical anomaly detection benchmarks.

desk verdict A useful empirical combination for medical AD, but the RSNA SOTA claim depends on a validation-selected aggregation that contradicts the main-text scoring rule. read the letter →

arxiv 2507.18481 v1 pith:F7LMV7PJ submitted 2025-07-24 cs.CV

Francesco Dalmonte , Emirhan Bayar , Emre Akbas , Mariana-Iuliana Georgescu This is my paper · ORCID
classification cs.CV
keywords medicalanomalydetectionunsupervisedQ-FormervisionfoundationmodelsDINOv2maskedautoencoderperceptualloss
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 classical autoencoder, modernized with three components, is a competitive unsupervised medical anomaly detector: a frozen vision foundation model as encoder, a Q-Former bottleneck that compresses multi-scale features into a fixed-length latent, and a perceptual loss computed in the feature space of a pretrained Masked Autoencoder. The claim matters because anomaly detection in medical images must usually work without labeled anomalies and without domain-specific fine-tuning. The authors report that the resulting framework, QFAE, reaches 94.3 AUROC on BraTS2021 brain MRI, 91.8 on RESC retinal OCT, and 83.8 on RSNA chest X-rays, matching or beating prior methods on all three benchmarks. Ablations attribute large gains to the Q-Former bottleneck and to the perceptual loss, indicating that neither component alone explains the result.

What carries the argument

The Q-Former bottleneck is the central adaptive component: a single transformer block with learnable query tokens that self-attend among themselves, cross-attend to the frozen encoder's projected multi-layer features as keys and values, and are passed through an MLP to produce a fixed-length latent. The number of queries controls the decoder's patch count, decoupling encoder and decoder output lengths and letting the decoder reconstruct at 8x8 patches. The second load-bearing mechanism is the Masked-Autoencoder perceptual loss, computed as a multi-scale cosine distance in feature space; it replaces pixel-level losses and, together with taking the maximum reconstruction error per layer when scoring, is what separates anomalies from normal tissue.

What would settle it

Train QFAE replacing the frozen DINOv2 encoder with a randomly initialized ViT of the same architecture, keeping the Q-Former, decoder, and MAE perceptual loss unchanged; if BraTS2021 AUROC stays near 94.3, the frozen foundation-model features are not load-bearing, while if it collapses, the transfer claim is confirmed. A second check is to run the exact RSNA configuration on a different chest X-ray set and see whether the image-level score distribution separates known pathologies from healthy controls.

Watch

Extended reading notes

Core claim

QFAE replaces the trainable encoder of a classical autoencoder with a frozen pretrained vision transformer, such as DINO, DINOv2, or OpenCLIP, feeds the resulting multi-scale features through a Q-Former bottleneck that compresses them into a fixed number of learnable queries, and reconstructs the image with a lightweight transformer decoder. Training is guided solely by a perceptual loss: the cosine distance between features of the input and the reconstruction, extracted from several layers of a frozen Masked Autoencoder at multiple patch sizes. At test time the same cosine distance, aggregated per-layer by taking the maximum spatial error and then averaging across layers, is the anomaly score. The paper reports that this setup reaches 94.3 AUROC on BraTS2021, 91.8 on RESC, and 83.8 on RSNA, matching or surpassing prior methods on each, with ablations showing the Q-Former bottleneck and the MAE perceptual loss each contribute large gains.

Load-bearing premise

The load-bearing premise is that features learned on natural images by DINOv2 and the Masked Autoencoder transfer to brain MRI, retinal OCT, and chest X-ray without any fine-tuning, and that cosine distance in MAE feature space between an image and its reconstruction reliably separates anomalies from healthy tissue across all three modalities.

Editorial extensions

If this is right

  • On the three BMAD benchmarks, QFAE matches or beats every compared method without any encoder fine-tuning or labeled anomalies.
  • Ablations attribute 12.9 AUROC points to the Q-Former bottleneck and further large gains to switching from pixel-level loss to MAE feature-space perceptual loss, so both components are needed.
  • Smaller decoder patch sizes (8x8) and larger perceptual patch sizes (32x32 and 56x56 combined) improve detection, indicating granularity choices matter on both the reconstruction and scoring sides.
  • Per-dataset evaluation tweaks, such as mean-then-max aggregation on chest X-rays with optical artifacts, change AUROC by several points, so the aggregation rule should be tuned to the modality.

Reading between the lines

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

  • A testable extension is multi-class anomaly detection: the paper names it as future work, and the fixed-length latent produced by Q-Former could plausibly carry a class label, but the paper does not test this.
  • The frozen-encoder design implies the framework's ceiling is set by the feature spaces of DINOv2 and MAE; newer foundation models pretrained on medical or multimodal data could be swapped in without retraining the rest, which the paper does not evaluate.
  • The sensitivity of RSNA results to switching the anomaly-score aggregation suggests the method may need a per-dataset aggregation recipe; whether a single universal aggregation exists is left open.
  • Because all reported gains come from datasets with relatively homogeneous normal images, a harder test would be a modality with large normal variation, such as full-body CT; the paper's stated limitation anticipates this.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes QFAE, an unsupervised medical anomaly detection autoencoder that combines a frozen pretrained ViT encoder (DINO, DINOv2, OpenCLIP), a Q-Former bottleneck producing a fixed-length latent sequence, a lightweight transformer decoder, and a perceptual loss computed from a frozen Masked Autoencoder. Anomaly scores are obtained from cosine distances between MAE features of the input and its reconstruction. Experiments on the BMAD benchmarks report AUROC values of 94.3 on BraTS2021, 91.8 on RESC, and 83.8 on RSNA, with additional LiverCT results in the supplement.

Significance. If the reported numbers hold under a clearly specified protocol, the paper would demonstrate that a frozen foundation-model encoder plus a Q-Former bottleneck and MAE-based perceptual loss is a competitive reconstruction-based anomaly detector across MRI, OCT, and X-ray modalities. The paper has clear strengths: code and models are released, results are reported as mean and standard deviation over five seeds, and the ablation study examines the encoder, loss, aggregation, decoder patch size, and perceptual-model patch size. The authors also include a limitations paragraph in the conclusions, which is a positive indication of balanced reporting. However, the headline SOTA claims are weakened by two issues: the BraTS2021 result is a statistical tie with P-VQ, and the RSNA result is produced by an anomaly-score aggregation different from the one defined in the main text and selected per dataset on the validation split. As a result, the paper currently supports 'strong empirical results' more strongly than 'a single unified framework with three state-of-the-art benchmarks.'

major comments (3)
  1. [Supplement Section 8 vs. Eq. (4), Table 5] The RSNA state-of-the-art result is not produced by the anomaly score defined in Eq. (4). Eq. (4) defines the image-level score as the mean, over feature layers, of each layer's maximum cosine distance (max-then-mean). Supplement Section 8 reports that this aggregation obtained 78.6 AUROC on the RSNA validation split, whereas the alternative mean-then-max aggregation obtained 84.3, and that the test AUROC of 83.8 in Table 5 was obtained with the latter. Thus the RSNA claim depends on a per-dataset, validation-selected anomaly-score function that the main text does not define. The paper should either incorporate this alternative aggregation into the method definition for RSNA, report results for the fixed score of Eq. (4) across all datasets, or clearly restate the RSNA result as coming from a variant evaluated under a separate validation-selected protocol. As written, the central claim that QFAE as specified achieves SOTA on RSNA is not established.
  2. [Section 4.4, Table 5, Abstract] The 'state-of-the-art' claim on BraTS2021 is a statistical tie rather than an improvement: QFAE achieves 94.3±0.18 and P-VQ achieves 94.3±0.23. The abstract and the contributions bullet state SOTA on BraTS2021, and Table 5's caption says the method 'outperform[s] all methods,' while Section 4.4 says only that the result is 'on par' with P-VQ. This inconsistency should be resolved by either performing and reporting a significance test or softening the abstract and contributions to 'matches the state of the art on BraTS2021.'
  3. [Supplement Tables 6, 7, 9, 10; Section 6.3] The evaluation configuration is not a single fixed protocol: perceptual patch sizes, feature layers, and the order of spatial/cross-feature aggregation change per dataset and were selected on validation splits (e.g., Eq. (4) is used for BraTS/RESC, mean-then-max for RSNA, and a different patch-size set for LiverCT). Validation-based tuning is not inherently illegitimate, but the main text presents Eq. (4) as the method's score while the supplement shows that different scores are used for different reported results. The paper should specify the model-selection protocol, report results for both the fixed configuration and the per-dataset configurations, and state in the main text which evaluation configuration produced each headline number. Without this, the claim that the framework generalizes across modalities is weaker than presented.
minor comments (5)
  1. [Supplement Section 6.3] The formula 'Apixel-max(j,k) = max_n (An(j,k))' contradicts the preceding sentence, which says the final pixel-level map is obtained by taking the mean across all resized anomaly maps; please correct the formula and the variable name.
  2. [Supplement Table 8] The first row reports '654.1' for the Main Config AUROC; this appears to be a typo for '65.4' and should be corrected.
  3. [Section 4.2 vs. Supplement Table 6] The main text says features are extracted from 'layers 20 and 22 of the ViT-L encoder and layers 8 and 10 of the ViT-B architecture,' while the supplement specifies '2nd and 4th to last blocks'; please align these descriptions for reproducibility.
  4. [Table 1 vs. Table 2a] Table 1 row 3 reports 86.8 AUROC for the LPerceptual configuration, while Table 2a reports 88.5 AUROC for the same loss setting; please state the exact configuration for each table or explain the discrepancy.
  5. [Figure 4 caption] The caption notes that MVFA-AD is few-shot rather than unsupervised; this qualification should also appear in the main text where MVFA-AD is compared with unsupervised methods.

Circularity Check

1 steps flagged · score 3.0 of 10

RSNA state-of-the-art number uses a validation-selected anomaly-score aggregation that differs from the paper's Eq. (4); otherwise the derivation is self-contained.

  1. fitted input called prediction [Supplementary Section 8 (Different Aggregation for Chest RSNA); cf. main-text Eq. (4) and Tables 7 and 10]
    "To mitigate this problem, we decided to experiment with different aggregation methods on the validation split of the Chest RSNA dataset. As an alternative, we first tried taking the mean value in the anomaly map from each location, and then taking the maximum across different layers. We observed an increase in AUROC from 78.6% to 84.3% on the validation split. Therefore, we decided to keep this approach and reported an AUROC of 83.8% on test set as in main Table 5."

    Eq. (4) defines the image-level score as max-then-mean: Ascore = (1/|I|) sum_i max(A_map,i). The RSNA configuration in Table 10 instead uses spatial aggregation 'Mean' and cross-feature 'Max', i.e. mean-then-max, a different scoring function. That function was chosen by measuring AUROC on the RSNA validation split (78.6 -> 84.3) and then carried to test. Thus the headline 'state-of-the-art on RSNA' (83.8) is not produced by the framework as specified in Eq. (4); it is produced by a scoring rule fitted to the validation split. The paper presents Eq. (4) as the score while reporting a per-dataset variant.

full rationale

No load-bearing circularity is found in the main derivation chain. The training loss (Eq. 2) and the anomaly score (Eqs. 3-4) are both cosine distances in Masked-AE feature space; this is the standard reconstruction-error paradigm, and whether it separates anomalies is an empirical claim evaluated on external test labels, not a tautology. The use of frozen DINOv2/DINO/MAE features, the Q-Former bottleneck, and the perceptual loss are defended with external citations or ablations, not with self-citations. The only self-citation (Ref. [17], an author's prior masked-autoencoder AD paper) appears in related work and is not used to justify the central architecture or to exclude alternatives. The one substantive concern is the RSNA anomaly-score aggregation: the main text defines max-then-mean in Eq. (4), while the supplementary reports that the RSNA state-of-the-art result uses a mean-then-max aggregation selected on the validation split. This makes the RSNA claim a fitted evaluation variant rather than a prediction of the method as specified. BraTS and RESC results are internally consistent with Eq. (4) and are compared to external baselines, so those claims retain independent content. The overall circularity score is therefore low (3/10): one fitted input affects one of three headline benchmarks, but the core framework is not circular.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The framework contributes no new entities. The main burden is the assumption that frozen natural-image features transfer to medical modalities, and the per-dataset evaluation configuration choices act as free parameters fitted to validation AUROC.

free parameters (6)
  • Q-Former query count = 784 (28x28)
    Set by the 8x8 output patch size; chosen based on validation experiments to control reconstruction granularity (Section 3.2, Table 2d).
  • Decoder output patch size = 8x8
    Ablated in Table 2d; 8x8 chosen because it yields the highest BraTS AUROC (93.0 vs 92.5 for 16, 91.1 for 32).
  • Perceptual loss training layers = blocks 16 and 20 of MAE ViT-L
    Chosen in Section 6.2; the Table 2c ablation shows 3 layers (11,15,19) give 93.0, but the default uses 2 layers, so the exact choice is not uniquely determined by the reported ablation.
  • Perceptual model patch sizes (training) = 32x32, 56x56
    Selected from Table 3, where this combination gives 94.4 on BraTS vs 93.0 for a smaller set; these are free choices tuned on validation.
  • Perceptual model patch sizes (evaluation) = 16x16, 32x32, 56x56 (BraTS/RESC/RSNA); 8x16 (LiverCT)
    Per-dataset evaluation configs in Supplement Tables 7, 9, 10; changed after observing validation AUROC, so they are fitted to the benchmark.
  • Anomaly score aggregation order = max-then-mean (main), mean-then-max (RSNA)
    Main text Eq. 4 uses max-then-mean; Supplement Section 8 changes to mean-then-max for RSNA after validation AUROC rose from 78.6 to 84.3.
assumptions (4)
  • domain assumption Frozen natural-image foundation models (DINOv2, MAE) provide feature representations that transfer to medical images without fine-tuning.
    The whole framework relies on this; Section 4.4 claims it, but no evidence is given beyond the benchmark results.
  • domain assumption Anomalies are harder to reconstruct than normal data in the chosen MAE feature space.
    Standard autoencoder assumption; stated in Section 3.1 but not proven for medical anomalies.
  • domain assumption The MAE perceptual loss preserves anomaly-relevant semantic structures.
    Used to justify the training objective in Section 3.2; relies on prior perceptual loss literature.
  • ad hoc to paper The evaluation configuration may be selected per dataset using validation performance.
    The method is defined differently for RSNA and LiverCT in the supplement, so the single-framework claim assumes this per-dataset tuning is acceptable.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Q-Former Autoencoder: A Modern Framework for Medical Anomaly Detection." pith.science (2026). https://pith.science/paper/F7LMV7PJ

@misc{pith2026250718481,
  author       = {Pith},
  title        = {Pith review of: Q-Former Autoencoder: A Modern Framework for Medical Anomaly Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F7LMV7PJ}},
  note         = {Machine review of arXiv:2507.18481}
}
read the original abstract

Anomaly detection in medical images is an important yet challenging task due to the diversity of possible anomalies and the practical impossibility of collecting comprehensively annotated data sets. In this work, we tackle unsupervised medical anomaly detection proposing a modernized autoencoder-based framework, the Q-Former Autoencoder, that leverages state-of-the-art pretrained vision foundation models, such as DINO, DINOv2 and Masked Autoencoder. Instead of training encoders from scratch, we directly utilize frozen vision foundation models as feature extractors, enabling rich, multi-stage, high-level representations without domain-specific fine-tuning. We propose the usage of the Q-Former architecture as the bottleneck, which enables the control of the length of the reconstruction sequence, while efficiently aggregating multiscale features. Additionally, we incorporate a perceptual loss computed using features from a pretrained Masked Autoencoder, guiding the reconstruction towards semantically meaningful structures. Our framework is evaluated on four diverse medical anomaly detection benchmarks, achieving state-of-the-art results on BraTS2021, RESC, and RSNA. Our results highlight the potential of vision foundation model encoders, pretrained on natural images, to generalize effectively to medical image analysis tasks without further fine-tuning. We release the code and models at https://github.com/emirhanbayar/QFAE.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 4 linked inside Pith

  1. [1]

    Ganomaly: Semi-supervised anomaly detection via adversarial training

    Samet Akcay, Amir Atapour-Abarghouei, and Toby P Breckon. Ganomaly: Semi-supervised anomaly detection via adversarial training. InAsian conference on computer vision, pages 622–637. Springer, 2018. 7, 15

    2018

  2. [2]

    The rsna-asnr-miccai brats 2021 benchmark on brain tumor segmentation and radiogenomic classification

    Ujjwal Baid, Satyam Ghodasara, Suyash Mohan, Michel Bilello, Evan Calabrese, Errol Colak, Keyvan Farahani, Jayashree Kalpathy-Cramer, Felipe C Kitamura, Sarthak Pati, et al. The rsna-asnr-miccai brats 2021 benchmark on brain tumor segmentation and radiogenomic classification. arXiv preprint arXiv:2107.02314, 2021. 2, 5, 6, 7, 8, 15

    arXiv 2021

  3. [3]

    Advancing the cancer genome atlas glioma mri collections with expert seg- mentation labels and radiomic features.Scientific data, 4(1): 1–13, 2017

    Spyridon Bakas, Hamed Akbari, Aristeidis Sotiras, Michel Bilello, Martin Rozycki, Justin S Kirby, John B Freymann, Keyvan Farahani, and Christos Davatzikos. Advancing the cancer genome atlas glioma mri collections with expert seg- mentation labels and radiomic features.Scientific data, 4(1): 1–13, 2017. 2, 5, 6, 7, 8, 15

    2017

  4. [4]

    Bmad: Benchmarks for medical anomaly detection

    Jinan Bao, Hanshi Sun, Hanqiu Deng, Yinsheng He, Zhaox- iang Zhang, and Xingyu Li. Bmad: Benchmarks for medical anomaly detection. In2024 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition Workshops (CVPRW), pages 4042–4053, 2024. 2, 5, 8, 14

    2024

  5. [5]

    Improving unsupervised defect seg- mentation by applying structural similarity to autoencoders

    Paul Bergmann, Sindy L ¨owe, Michael Fauser, David Sattleg- ger, and Carsten Steger. Improving unsupervised defect seg- mentation by applying structural similarity to autoencoders. InProceedings of the 14th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2019) - Volume 5: VISAPP, pages 37...

    2019

  6. [6]

    The liver tumor segmentation benchmark (lits).arXiv preprint arXiv:1901.04056, 2019

    Patrick Bilic, Patrick Ferdinand Christ, Eugene V orontsov, Grzegorz Chlebus, Hao Chen, Qi Dou, Chi-Wing Fu, Xiao Han, Pheng-Ann Heng, J ¨urgen Hesser, et al. The liver tumor segmentation benchmark (lits).arXiv preprint arXiv:1901.04056, 2019. 2, 5, 13, 15

    arXiv 1901

  7. [7]

    The liver tumor segmentation benchmark (lits).Medical Image Analysis, 84:102680, 2023

    Patrick Bilic, Patrick Christ, Hongwei Bran Li, Eugene V orontsov, Avi Ben-Cohen, Georgios Kaissis, Adi Szeskin, Colin Jacobs, Gabriel Efrain Humpire Mamani, Gabriel Chartrand, et al. The liver tumor segmentation benchmark (lits).Medical Image Analysis, 84:102680, 2023. 14

    2023

  8. [8]

    Rethinking Au- toencoders for Medical Anomaly Detection from A Theoreti- cal Perspective

    Yu Cai, Hao Chen, and Kwang-Ting Cheng. Rethinking Au- toencoders for Medical Anomaly Detection from A Theoreti- cal Perspective. InMedical Image Computing and Computer Assisted Intervention – MICCAI 2024: 27th International Conference, Marrakesh, Morocco, October 6–10, 2024, Pro- ceedings, Part XI, pages 544–554, Berlin, Heidelberg, 2024. Springer-Verlag. 2

    2024

Show all 27 references
  1. [9]

    MedIAnomaly: A comparative study of anomaly detection in medical images, 2025

    Yu Cai, Weiwen Zhang, Hao Chen, and Kwang-Ting Cheng. MedIAnomaly: A comparative study of anomaly detection in medical images, 2025. arXiv:2404.04518 [cs]. 2

    2025 arXiv

  2. [10]

    Emerg- ing properties in self-supervised vision transformers

    Mathilde Caron, Hugo Touvron, Ishan Misra, Herv ´e Jegou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. Emerg- ing properties in self-supervised vision transformers. In 2021 IEEE/CVF International Conference on Computer Vi- sion (ICCV), pages 9630–9640, 2021. 2, 3, 4, 5, 6, 8

    2021

  3. [11]

    Utrad: Anomaly detection and localization with u-transformer.Neural Networks, 147:53–62, 2022

    Liyang Chen, Zhiyuan You, Nian Zhang, Juntong Xi, and Xinyi Le. Utrad: Anomaly detection and localization with u-transformer.Neural Networks, 147:53–62, 2022. 7, 15

    2022

  4. [12]

    Padim: a patch distribution modeling framework for anomaly detection and localization

    Thomas Defard, Aleksandr Setkov, Angelique Loesch, and Romaric Audigier. Padim: a patch distribution modeling framework for anomaly detection and localization. InInter- national Conference on Pattern Recognition, pages 475–489. Springer, 2021. 7, 15

    2021

  5. [13]

    Anomaly detection via reverse distillation from one-class embedding

    Hanqiu Deng and Xingyu Li. Anomaly detection via reverse distillation from one-class embedding. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9737–9746, 2022. 7, 15

    2022

  6. [14]

    Anomaly Detection via Re- verse Distillation from One-Class Embedding

    Hanqiu Deng and Xingyu Li. Anomaly Detection via Re- verse Distillation from One-Class Embedding. In2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 9727–9736, New Orleans, LA, USA, 2022. IEEE. 3

    2022

  7. [15]

    An image is worth 16x16 words: Transformers for image recognition at scale.ArXiv, abs/2010.11929, 2020

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition ...

    2010 arXiv

  8. [16]

    Zero-shot out-of-distribution detection based on the pre-trained model clip.Proceedings of the AAAI Conference on Artificial Intelligence, 36(6):6568–6576, 2022

    Sepideh Esmaeilpour, Bing Liu, Eric Robertson, and Lei Shu. Zero-shot out-of-distribution detection based on the pre-trained model clip.Proceedings of the AAAI Conference on Artificial Intelligence, 36(6):6568–6576, 2022. 2

    2022

  9. [17]

    Masked Autoencoders for Un- supervised Anomaly Detection in Medical Images.Procedia Computer Science, 225:969–978, 2023

    Mariana-Iuliana Georgescu. Masked Autoencoders for Un- supervised Anomaly Detection in Medical Images.Procedia Computer Science, 225:969–978, 2023. 3

    2023

  10. [18]

    Anomalygpt: Detecting in- dustrial anomalies using large vision-language models

    Zhaopeng Gu, Bingke Zhu, Guibo Zhu, Yingying Chen, Ming Tang, and Jinqiao Wang. Anomalygpt: Detecting in- dustrial anomalies using large vision-language models. In AAAI Conference on Artificial Intelligence, 2023. 3

    2023

  11. [19]

    Cflow-ad: Real-time unsupervised anomaly detection with localization via conditional normalizing flows

    Denis Gudovskiy, Shun Ishizaka, and Kazuki Kozuka. Cflow-ad: Real-time unsupervised anomaly detection with localization via conditional normalizing flows. InProceed- ings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pages 98–107, 2022. 7, 15

    2022

  12. [20]

    ReContrast: domain-specific anomaly detection via con- trastive reconstruction

    Jia Guo, Shuai Lu, Lize Jia, Weihang Zhang, and Huiqi Li. ReContrast: domain-specific anomaly detection via con- trastive reconstruction. InProceedings of the 37th Inter- national Conference on Neural Information Processing Sys- tems, pages 10721–10740, Red Hook, NY , USA, 202...

    2023

  13. [21]

    Encoder-Decoder Contrast for Unsupervised Anomaly De- tection in Medical Images.IEEE Transactions on Medical Imaging, 43(3):1102–1112, 2024

    Jia Guo, Shuai Lu, Lize Jia, Weihang Zhang, and Huiqi Li. Encoder-Decoder Contrast for Unsupervised Anomaly De- tection in Medical Images.IEEE Transactions on Medical Imaging, 43(3):1102–1112, 2024. 3

    2024

  14. [22]

    Girshick

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Pi- otr Doll’ar, and Ross B. Girshick. Masked autoencoders are scalable vision learners.2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 15979–15988, 2021. 2, 3, 4, 5, 6

    2022

  15. [23]

    Automated seg- mentation of macular edema in oct using deep neural net- works.Medical image analysis, 55:216–227, 2019

    Junjie Hu, Yuanyuan Chen, and Zhang Yi. Automated seg- mentation of macular edema in oct using deep neural net- works.Medical image analysis, 55:216–227, 2019. 2, 5, 7, 8, 15

    2019

  16. [24]

    Self-supervised masking for unsupervised 9 anomaly detection and localization.IEEE Transactions on Multimedia, 2022

    Chaoqin Huang, Qinwei Xu, Yanfeng Wang, Yu Wang, and Ya Zhang. Self-supervised masking for unsupervised 9 anomaly detection and localization.IEEE Transactions on Multimedia, 2022. 2

    2022

  17. [25]

    Chaoqin Huang, Aofan Jiang, Jinghao Feng, Ya Zhang, Xin- chao Wang, and Yanfeng Wang. Adapting visual-language models for generalizable anomaly detection in medical im- ages.2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 11375–11385, 2024. 3, 5, 8

    2024

  18. [26]

    Winclip: Zero- /few-shot anomaly classification and segmentation.2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 19606–19616, 2023

    Jongheon Jeong, Yang Zou, Taewan Kim, Dongqing Zhang, Avinash Ravichandran, and Onkar Dabeer. Winclip: Zero- /few-shot anomaly classification and segmentation.2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 19606–19616, 2023. 3

    2023

  19. [27]

    Percep- tual losses for real-time style transfer and super-resolution

    Justin Johnson, Alexandre Alahi, and Li Fei-Fei. Percep- tual losses for real-time style transfer and super-resolution. InComputer Vision – ECCV 2016, pages 694–711, Cham,

    2016

Pith tools

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