Unsupervised Out-of-Distribution Detection by Maximum Classifier Discrepancy

Kiyoharu Aizawa; Qing Yu

REVIEW 4 major objections 3 minor 33 references

Unsupervised Out-of-Distribution Detection by Maximum Classifier Discrepancy

T0 review · 4 major / 3 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a two-head classifier fine-tuned on unlabeled test data, including out-of-distribution images, separates OOD from in-distribution inputs with AUROC above 99% on most benchmarks.

desk verdict The proposed transductive OOD detector is clearly explained and the idea is worth exploring, but the reported near-perfect results are an artifact of fine-tuning on the same test images used for evaluation. read the letter →

arxiv 1908.04951 v1 pith:DOFBEBK6 submitted 2019-08-14 cs.CV

Qing Yu , Kiyoharu Aizawa This is my paper · ORCID

classification cs.CV

keywords out-of-distributiondetectionmaximumclassifierdiscrepancyunsupervisedfine-tuningtwo-headsoftmaxtransductivelearningCIFAR-100TinyImageNet

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that out-of-distribution (OOD) inputs can be detected almost perfectly by fine-tuning a two-head classifier on unlabeled test data, maximizing the disagreement between the two heads. The authors propose a network with one shared feature extractor and two classifiers that agree on in-distribution (ID) images but are pushed to disagree on anything else. The driving idea is that the discrepancy between the two classifiers' softmax outputs marks whether an input lies outside the ID support. On benchmarks such as CIFAR-100 as ID with TinyImageNet, LSUN, or iSUN as OOD, the method reports AUROC values of 99.6–100.0, far above prior detectors. The practical stake is that OOD detection could be improved cheaply by exploiting unlabeled data available at test time.

What carries the argument

The central object is a two-head CNN: a shared feature extractor plus two classifier heads, $F_1$ and $F_2$, trained with a margin-based discrepancy loss $d(p_1, p_2) = H(p_1) - H(p_2)$. The margin $m$ in the unsupervised loss prevents overfitting by stopping optimization once the average discrepancy exceeds $m$. The machinery works by using the unlabeled set $X_{ul}$ (the test split) to maximize the gap between the decision boundaries of the two heads, effectively relocating OOD samples outside the ID manifold. At test time, the L1 distance between the two softmax outputs is the OOD score.

What would settle it

Evaluate on OOD classes that never appear in the unlabeled fine-tuning set (e.g., use only half the TinyImageNet classes in $X_{ul}$ and test on the other half); if AUROC drops substantially from the near-perfect values, the detector is exploiting the specific OOD images rather than a general notion of being out-of-distribution.

Watch

Extended reading notes

Core claim

The central discovery is that maximizing the discrepancy between two classifiers on unlabeled data, most of it drawn from the test split, separates ID from OOD samples. The paper defines the discrepancy loss as the difference in entropy between the two softmax outputs, $H(p_1) - H(p_2)$, and during fine-tuning alternately trains the network to classify labeled ID images correctly and to maximize this discrepancy on unlabeled images. This pushes OOD images toward high-entropy predictions from one head and low-entropy from the other, while ID images remain tightly classified by both. At inference, an input is flagged OOD if the L1 distance between the two softmax vectors exceeds a threshold. The paper demonstrates near-perfect separation on standard benchmark pairs and on simulated food and fashion applications, with a simple interpretable threshold near 1.0.

Load-bearing premise

The method uses the test split, including the OOD test images, as the unlabeled data for fine-tuning, and then evaluates on that same split, so the near-perfect scores may reflect seeing the test OOD data rather than a generalizable OOD detector.

Editorial extensions

If this is right

OOD detection can be performed without any labeled OOD samples, using only unlabeled data, which is easy to obtain.
The method works even when the unlabeled set contains few OOD samples (e.g., 500 OOD among 9,000 ID), as shown in ablation studies.
The method generalizes when the OOD dataset used in fine-tuning differs from the OOD dataset used for evaluation (e.g., fine-tune on TinyImageNet-crop, test on LSUN-crop).
The approach can be dropped onto existing architectures (DenseNet, Wide ResNet) with minimal modification.
A single threshold near 1.0 separates ID from OOD, unlike earlier methods requiring careful threshold tuning.

Reading between the lines

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

A natural test is to fine-tune on an unlabeled set containing only ID images (no OOD at all) and then evaluate on OOD test images; a large AUROC drop would indicate the method relies on seeing OOD samples during fine-tuning.
The transductive setup means the reported near-perfect AUROC should be read as an upper bound for a system that must generalize to future OOD inputs; a fair comparison with methods that do not see test data would require a separate unlabeled set disjoint from the evaluation OOD set.
The discrepancy principle might transfer to non-image modalities where unlabeled data is abundant, though the paper only evaluates images; this remains an open empirical question.

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.

Desk Editor's Note

read the letter

Colleague,

The paper proposes a two-head classifier network for out-of-distribution detection: after supervised pretraining on labeled ID data, it fine-tunes with an unsupervised loss that maximizes the discrepancy between two classifiers on unlabeled data. That unlabeled data is the test split itself. The authors are upfront about this: Section 3.4 says "we use the test data as the unlabeled data," and Section 4.1.3 says the remaining test images (ID plus OOD) were used "as Xul for unsupervised training and evaluation." This means the model is adapted directly on the images it is then scored on. The AUROC numbers in Table 2 (99.6–100.0) are fitted numbers, not estimates of how the detector behaves on new OOD inputs. The comparison to ODIN and ELOC, which never see the test set, is invalid.

That is the main problem, and it is not a subtle one. The paper's own description makes it explicit. The phrase "totally unseen samples were included during evaluation" in the ablation section is inaccurate: samples were unseen in the sense of having no labels, but they were seen by the optimiser during fine-tuning. Table 4 varies the OOD source, but ID test images are still shared between Xul and evaluation, so ID-side leakage remains.

There is something worth building on. Using unlabeled data for OOD detection is a sensible direction, and borrowing the maximum classifier discrepancy from Saito et al. is a legitimate adaptation. The writing is clear, the ablation on Xul size and composition is informative, and the real-world simulations are a nice extra. But the central evidence does not survive a re-run under a proper protocol: you would need to fine-tune on one unlabeled batch and evaluate on a disjoint batch (or at least hold out part of the unlabeled set). The 5% classification accuracy drop noted in the limitation section also deserves attention.

For a peer review decision: I would not accept these results as evidence of a state-of-the-art OOD detector. The paper needs a major revision with a clean evaluation protocol before it can be fairly judged. If the transductive framing is intentional, the authors should argue why a detector that has already seen the test batch is useful, and then evaluate on a different batch.

This belongs in the "interesting idea, invalid execution" pile. I would not cite the numbers as they stand.

Referee Report

4 major / 3 minor

Summary. The paper proposes a two-head CNN with a shared feature extractor and two classifiers for out-of-distribution (OOD) detection. The method first trains the network on labeled in-distribution (ID) data, then fine-tunes it using unlabeled data while maximizing the discrepancy between the two classifiers' softmax outputs, defined as H(p1) - H(p2). At inference, a sample is declared OOD when the L1 distance between the two classifiers' probability vectors exceeds a threshold. The authors report near-perfect AUROC values on CIFAR-10/CIFAR-100 benchmarks, real-world food and fashion simulations, and ablation studies. The central claim is that this 'unsupervised' fine-tuning on test data is both legitimate and highly effective, significantly outperforming prior methods that only use labeled ID training data.

Significance. If the empirical evaluation were sound, the idea of using two classifiers and maximizing their discrepancy on unlabeled data would be a meaningful contribution, and the reported AUROC values of 99.6-100.0 on several benchmarks would be a large improvement over prior art. The paper also provides a clear description of the method, extensive ablations, and two real-world simulation cases. However, the evaluation protocol is fundamentally invalid: the unlabeled data used for fine-tuning is the same test data used for evaluation, including the OOD test images. The loss in Eq. (5) explicitly maximizes the discrepancy on those exact test images, so the results are fitted, not predicted. This invalidates the head-to-head comparison with ODIN and ELOC, which do not see the test data during training, and makes the reported performance uninterpretable as a measure of OOD detection generalization.

major comments (4)

[Section 4.1.3 and Section 3.4] The evaluation is contaminated because the unlabeled set Xul used for fine-tuning is the same set on which the method is evaluated. Section 4.1.3 states that the remaining test images, including OOD images, were used as Xul for unsupervised training and evaluation, and Section 3.4 says 'we use the test data as the unlabeled data.' Since Step B (Eq. (5)) directly maximizes the discrepancy on these exact test images, the AUROC values in Table 2 are fitted values, not measures of detection performance on unseen OOD inputs. The comparison with ODIN and ELOC, which do not use test data during training, is therefore not a fair comparison.
[Section 4.1.7, Table 3 and Table 4] The ablation studies do not resolve the leakage problem. In Table 3, the 9,000 ID and 9,000 OOD samples used for testing are the same samples that constitute Xul, so the claim that 'totally unseen samples were included during evaluation' is misleading. In Table 4, even when the OOD dataset used for testing is different from the OOD dataset in Xul, the ID test images are still the same as the ID images in Xul, so there remains ID-side leakage. No setting in the paper evaluates on a test set disjoint from Xul.
[Section 3.5 and Eq. (6)] The inference rule uses an L1-distance threshold delta, but the paper does not explain how delta is selected or whether it is tuned on the validation split. If delta is tuned on the validation set, it is not independent of the fine-tuned model because the model was trained on Xul, which shares the same distribution as the validation set. More importantly, the evaluation on Xul means that the threshold and the model are both adapted to the evaluation data, further inflating the reported metrics.
[Section 4.1.5 and Figure 5] The histogram in Figure 5a, showing that a simple threshold of 1.0 separates ID and OOD samples, is computed on Xul, the same data used for fine-tuning. This visualization therefore demonstrates that the model has fitted the specific OOD test set, not that it has learned a generalizable separation. The same issue applies to the claim that the two classifiers' maximum softmax scores for ID and OOD are cleanly separated after fine-tuning.

minor comments (3)

[Section 3.3, Eq. (1)] The discrepancy term H(p1) - H(p2) is asymmetric with respect to the two classifiers, but the training procedure treats F1 and F2 symmetrically in Eq. (2). The paper should explain why this asymmetry is justified and whether the roles of F1 and F2 matter.
[Section 4.1.1] The description of the fair-comparison modification for ODIN and ELOC is unclear: the paper says 'we used two classifiers and calculated the average score of these two classifiers as final output in the other methods,' but it is not specified how ODIN's temperature scaling and input preprocessing are applied in this ensemble setting.
[Figure 4] Figure 4 is described as a histogram of 'the discrepancy (L1 distance)' between the two classifiers' outputs, but Section 3.3 defines discrepancy as H(p1) - H(p2). The paper should clarify which quantity is plotted in Figure 4.

Circularity Check

2 steps flagged · score 8.0 of 10

Reported AUROC is a transductive fit: the model is fine-tuned on the same test images (Xul) that are later scored, so the central 'prediction' is forced by construction.

fitted input called prediction [Section 3.4 (Training Procedure), Section 4.1.3 (Out-of-Distribution), Eq. (5), Table 2]
""In principle, we use the test data as the unlabeled data." ... "the remaining test images containing unlabeled ID or OOD samples were used as Xul for unsupervised training and evaluation." Eq. (5): "Lunsup = max (m − (1/|Xul|) Σ_{xul∈Xul} d(p1(y|xul), p2(y|xul)), 0).""

The discrepancy loss in Eq. (5) is minimized by increasing the average classifier discrepancy on Xul, and Xul is exactly the remaining test set on which Table 2 computes AUROC, FPR, and detection error. Thus the optimization directly pushes the OOD test images' discrepancy scores above the margin, and the reported 99.6–100.0 AUROC measures how well the model fit the evaluation set, not how well it predicts OOD on unseen inputs. Since ODIN and ELOC are never fine-tuned on test images, the comparison is not on equal footing, and the headline 'prediction' reduces to a fitted outcome.
fitted input called prediction [Section 4.1.7 (Ablation Studies), Table 3]
""Even when 9,000 ID samples and 500 OOD samples are included in Xul, our method still have better performance than [16, 26]... Please notice that we used all 9,000 ID samples and 9,000 OOD samples for testing, which means totally unseen samples were included during evaluation.""

In the ablation row with 9,000 ID and 9,000 OOD samples in Xul, the paper then uses all 9,000 ID and 9,000 OOD samples for testing, so the evaluation images coincide with the unsupervised training images. The statement that 'totally unseen samples were included during evaluation' is contradicted by the table, which shows the tested samples were used as Xul. The low detection error in that setting is therefore a transductive fit on the evaluation set, not evidence of generalization to unseen OOD data.

full rationale

The core methodological contribution, maximizing classifier discrepancy on unlabeled data, is a reasonable training objective whose internal math is self-consistent. The circularity is not in the loss derivation but in the evaluation protocol: Section 3.4 says 'we use the test data as the unlabeled data,' and Section 4.1.3 confirms that the remaining test images are 'used as Xul for unsupervised training and evaluation.' Equation (5) then optimizes the discrepancy exactly on those same ID and OOD test images, and Table 2 reports AUROC on them. This makes the central claim, state-of-the-art OOD detection, a fitted-input-called-prediction result rather than a predictive one. Table 4 provides some partial relief by testing on OOD datasets different from those in Xul in two rows, but the ID test images used for evaluation are still drawn from the same Xul pool, so ID-side leakage remains. There is no self-citation chain or imported uniqueness theorem; the problem is entirely the same-split train/evaluation design. The paper even acknowledges the split in its data-preparation text, so the reduction is explicit and quotable. The appropriate score is 8: the headline numbers are forced by construction with respect to the evaluation set, while the underlying discrepancy mechanism retains independent methodological content.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The central contribution is a training procedure, not a new physical entity. The main free parameters are the margin, learning rate, and epochs. The key unstated assumption is that using the test data for fine-tuning is a valid way to evaluate a detector; this assumption is the source of the inflated results.

free parameters (4)

margin m = 1.2
The margin in the unsupervised loss (Eq. 5) is set to 1.2 by hand. It controls how much discrepancy is enforced and is not derived from first principles.
fine-tuning learning rate = 0.1
The learning rate for the fine-tuning steps is a hyperparameter chosen by the authors (Section 4.1.1).
number of fine-tuning epochs = 10
The network is fine-tuned for 10 epochs, a hand-chosen stopping point.
inference threshold delta = threshold swept for metrics; simple threshold 1.0 visualized
The detection threshold in Eq. (6) is not fixed for the reported metrics: AUROC and detection error sweep over thresholds. A 'simple threshold 1.0' is shown in Figure 5a, but it is not used for the main reported numbers.

assumptions (3)

domain assumption The unlabeled data Xul is drawn from the same distribution as the test data, and using it for fine-tuning does not invalidate the evaluation.
The paper assumes that transductive use of the test set is legitimate for measuring OOD detection performance. This assumption is load-bearing and, for standard inductive evaluation, false.
domain assumption Two classifiers trained on ID data with random initializations will have different decision boundaries for OOD samples but agree on ID samples.
The method relies on the empirical observation in Figure 4 that OOD samples show larger L1 discrepancy after supervised pre-training. This is not proven theoretically.
ad hoc to paper Maximizing the discrepancy loss (H(p1)-H(p2)) on unlabeled data pushes OOD samples outside the support of ID samples without harming ID classification too much.
The paper introduces this as the mechanism for OOD detection (Section 3.3). It is a modeling choice that is validated only empirically, and it is applied directly to the evaluation data, making the validation circular.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Unsupervised Out-of-Distribution Detection by Maximum Classifier Discrepancy." pith.science (2026). https://pith.science/paper/DOFBEBK6

@misc{pith2026190804951,
  author       = {Pith},
  title        = {Pith review of: Unsupervised Out-of-Distribution Detection by Maximum Classifier Discrepancy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DOFBEBK6}},
  note         = {Machine review of arXiv:1908.04951}
}

read the original abstract

Since deep learning models have been implemented in many commercial applications, it is important to detect out-of-distribution (OOD) inputs correctly to maintain the performance of the models, ensure the quality of the collected data, and prevent the applications from being used for other-than-intended purposes. In this work, we propose a two-head deep convolutional neural network (CNN) and maximize the discrepancy between the two classifiers to detect OOD inputs. We train a two-head CNN consisting of one common feature extractor and two classifiers which have different decision boundaries but can classify in-distribution (ID) samples correctly. Unlike previous methods, we also utilize unlabeled data for unsupervised training and we use these unlabeled data to maximize the discrepancy between the decision boundaries of two classifiers to push OOD samples outside the manifold of the in-distribution (ID) samples, which enables us to detect OOD samples that are far from the support of the ID samples. Overall, our approach significantly outperforms other state-of-the-art methods on several OOD detection benchmarks and two cases of real-world simulation.

Figures

Figures reproduced from arXiv: 1908.04951 by the authors.

**Figure 2.** Comparison of previous and the proposed OOD [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** Since OOD samples are not clearly categorized [PITH_FULL_IMAGE:figures/full_fig_p002_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Histogram of the discrepancy (L1 distance) be [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]

**Figure 3.** Figure 3: Fine-tuning steps of our method. Our network [PITH_FULL_IMAGE:figures/full_fig_p004_3.png]

**Figure 5.** Figure 5: The visualization of the result. TinyImageNet-resize, LSUN-resize and iSUN, which contain the images with full objects as opposed to the cropped parts of objects, are considered more difficult to detect. Our proposal shows highly accurate results on these more challeng…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 21 canonical work pages

[1]

Towards open world recognition

Abhijit Bendale and Terrance Boult. Towards open world recognition. In CVPR, 2015

2015
[2]

Towards open set deep networks

Abhijit Bendale and Terrance E Boult. Towards open set deep networks. In CVPR, 2016

2016
[3]

Food-101 – mining discriminative components with random forests

Lukas Bossard, Matthieu Guillaumin, and Luc Van Gool. Food-101 – mining discriminative components with random forests. In ECCV, 2014

work page 2014
[4]

Imagenet: A large-scale hierarchical image database

Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In CVPR, 2009

2009
[5]

Bert: Pre-training of deep bidirectional transformers for language understanding

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018

arXiv 2018
[6]

Generative openmax for multi-class open set clas- siﬁcation

Zongyuan Ge, Sergey Demyanov, Zetao Chen, and Rahil Garnavi. Generative openmax for multi-class open set clas- siﬁcation. In BMVC, 2017

work page 2017
[7]

Generative adversarial nets

Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. In NIPS, 2014

work page 2014
[8]

Deep residual learning for image recognition

Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In CVPR, 2016

2016

Show all 33 references

[9]

A baseline for detect- ing misclassiﬁed and out-of-distribution examples in neural networks

Dan Hendrycks and Kevin Gimpel. A baseline for detect- ing misclassiﬁed and out-of-distribution examples in neural networks. In ICLR, 2017

2017
[10]

Densely connected convolutional net- works

Gao Huang, Zhuang Liu, Laurens Van Der Maaten, and Kil- ian Q Weinberger. Densely connected convolutional net- works. In CVPR, 2017

2017
[11]

Delta divergence: A novel de- cision cognizant measure of classiﬁer incongruence

Josef Kittler and Cemre Zor. Delta divergence: A novel de- cision cognizant measure of classiﬁer incongruence. IEEE Transactions on Cybernetics, 2019

2019
[12]

Error sensitivity analysis of delta diver- gence - a novel measure for classiﬁer incongruence detec- tion

Josef Kittler, Cemre Zor, Ioannis Kaloskampis, Yulia Hicks, and Wenwu Wang. Error sensitivity analysis of delta diver- gence - a novel measure for classiﬁer incongruence detec- tion. Pattern Recognition, 2017

2017
[13]

Learning multiple layers of features from tiny images

Alex Krizhevsky. Learning multiple layers of features from tiny images. Technical report, 2009

2009
[14]

Training conﬁdence-calibrated classiﬁers for detecting out- of-distribution samples

Kimin Lee, Honglak Lee, Kibok Lee, and Jinwoo Shin. Training conﬁdence-calibrated classiﬁers for detecting out- of-distribution samples. In ICLR, 2018

2018
[15]

A simple uniﬁed framework for detecting out-of-distribution samples and adversarial attacks

Kimin Lee, Kibok Lee, Honglak Lee, and Jinwoo Shin. A simple uniﬁed framework for detecting out-of-distribution samples and adversarial attacks. In NIPS, 2018

2018
[16]

Shiyu Liang, Yixuan Li, and R. Srikant. Enhancing the re- liability of out-of-distribution image detection in neural net- works. In ICLR, 2018

2018
[17]

Sphereface: Deep hypersphere embedding for face recognition

Weiyang Liu, Yandong Wen, Zhiding Yu, Ming Li, Bhiksha Raj, and Le Song. Sphereface: Deep hypersphere embedding for face recognition. In CVPR, 2017

2017
[18]

Deepfashion: Powering robust clothes recognition and retrieval with rich annotations

Ziwei Liu, Ping Luo, Shi Qiu, Xiaogang Wang, and Xiaoou Tang. Deepfashion: Powering robust clothes recognition and retrieval with rich annotations. In CVPR, 2016

2016
[19]

Automatic differentiation in pytorch

Adam Paszke, Sam Gross, Soumith Chintala, Gregory Chanan, Edward Yang, Zachary DeVito, Zeming Lin, Al- ban Desmaison, Luca Antiga, and Adam Lerer. Automatic differentiation in pytorch. In NIPS-W, 2017

2017
[20]

Quintanilha, Roberto de M

Igor M. Quintanilha, Roberto de M. E. Filho, Jos Lezama, Mauricio Delbracio, and Leonardo O. Nunes. Detecting out- of-distribution samples using low-order deep features statis- tics. In Submitted to ICLR, 2019

2019
[21]

Rudd, Lalit P

Ethan M. Rudd, Lalit P. Jain, Walter J. Scheirer, and Ter- rance E. Boult. The extreme value machine. IEEE Transac- tions on Pattern Analysis and Machine Intelligence, 2018

2018
[22]

Maximum classiﬁer discrepancy for unsuper- vised domain adaptation

Kuniaki Saito, Kohei Watanabe, Yoshitaka Ushiku, and Tat- suya Harada. Maximum classiﬁer discrepancy for unsuper- vised domain adaptation. In CVPR, 2018

2018
[23]

Toward open set recogni- tion

Walter J Scheirer, Anderson de Rezende Rocha, Archana Sapkota, and Terrance E Boult. Toward open set recogni- tion. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013

2013
[24]

Prob- ability models for open set recognition

Walter J Scheirer, Lalit P Jain, and Terrance E Boult. Prob- ability models for open set recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014

2014
[25]

Out-of- distribution detection using multiple semantic label represen- tations

Gabi Shalev, Yossi Adi, and Joseph Keshet. Out-of- distribution detection using multiple semantic label represen- tations. In NIPS, 2018

2018
[26]

Out-of- distribution detection using an ensemble of self supervised leave-out classiﬁers

Apoorv Vyas, Nataraj Jammalamadaka, Xia Zhu, Dipankar Das, Bharat Kaul, and Theodore L Willke. Out-of- distribution detection using an ensemble of self supervised leave-out classiﬁers. In ECCV, 2018

2018
[27]

Be- yond novelty detection: Incongruent events, when general and speciﬁc classiﬁers disagree

Daphna Weinshall, Alon Zweig, Hynek Hermansky, Ste- fan Kombrink, Frank W Ohl, J ¨orn Anem¨uller, J¨org-Hendrik Bach, Luc Van Gool, Fabian Nater, Tomas Pajdla, et al. Be- yond novelty detection: Incongruent events, when general and speciﬁc classiﬁers disagree. IEEE Transaction...

2012
[28]

Sun database: Large-scale scene recognition from abbey to zoo

Jianxiong Xiao, James Hays, Krista A Ehinger, Aude Oliva, and Antonio Torralba. Sun database: Large-scale scene recognition from abbey to zoo. In CVPR, 2010

2010
[29]

Turkergaze: Crowdsourcing saliency with webcam based eye tracking

Pingmei Xu, Krista A Ehinger, Yinda Zhang, Adam Finkel- stein, Sanjeev R Kulkarni, and Jianxiong Xiao. Turkergaze: Crowdsourcing saliency with webcam based eye tracking. arXiv preprint arXiv:1504.06755, 2015

2015 arXiv
[30]

Classiﬁcation- reconstruction learning for open-set recognition

Ryota Yoshihashi, Wen Shao, Rei Kawakami, Shaodi You, Makoto Iida, and Takeshi Naemura. Classiﬁcation- reconstruction learning for open-set recognition. In CVPR, 2019

2019
[31]

Construction of a large- scale image dataset using deep learning with humans in the loop

Fisher Yu, Ari Seff, Yinda Zhang, Shuran Song, Thomas Funkhouser, and Jianxiong Xiao. Construction of a large- scale image dataset using deep learning with humans in the loop. arXiv preprint arXiv: 1506.03365, 2015

2015 arXiv
[32]

Wide residual net- works

Sergey Zagoruyko and Nikos Komodakis. Wide residual net- works. In BMVC, 2016

2016
[33]

Understanding deep learning re- quires rethinking generalization

Chiyuan Zhang, Samy Bengio, Moritz Hardt, Benjamin Recht, and Oriol Vinyals. Understanding deep learning re- quires rethinking generalization. In ICLR, 2017. 9

2017

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Towards open world recognition

Abhijit Bendale and Terrance Boult. Towards open world recognition. In CVPR, 2015

2015

[2] [2]

Towards open set deep networks

Abhijit Bendale and Terrance E Boult. Towards open set deep networks. In CVPR, 2016

2016

[3] [3]

Food-101 – mining discriminative components with random forests

Lukas Bossard, Matthieu Guillaumin, and Luc Van Gool. Food-101 – mining discriminative components with random forests. In ECCV, 2014

work page 2014

[4] [4]

Imagenet: A large-scale hierarchical image database

Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In CVPR, 2009

2009

[5] [5]

Bert: Pre-training of deep bidirectional transformers for language understanding

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018

arXiv 2018

[6] [6]

Generative openmax for multi-class open set clas- siﬁcation

Zongyuan Ge, Sergey Demyanov, Zetao Chen, and Rahil Garnavi. Generative openmax for multi-class open set clas- siﬁcation. In BMVC, 2017

work page 2017

[7] [7]

Generative adversarial nets

Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. In NIPS, 2014

work page 2014

[8] [8]

Deep residual learning for image recognition

Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In CVPR, 2016

2016