The DOI matches its Crossref record; 27 of the 32 references it cites were found in a registry or an index.
Generative OpenMax for Multi-Class Open Set Classification
Open accessAbstract
We present a conceptually new and flexible method for multi-class open set classification. Unlike previous methods where unknown classes are inferred with respect to the feature or decision distance to the known classes, our approach is able to provide explicit modelling and decision score for unknown classes. The proposed method, called Gener- ative OpenMax (G-OpenMax), extends OpenMax by employing generative adversarial networks (GANs) for novel category image synthesis. We validate the proposed method on two datasets of handwritten digits and characters, resulting in superior results over previous deep learning based method OpenMax Moreover, G-OpenMax provides a way to visualize samples representing the unknown classes from open space. Our simple and effective approach could serve as a new direction to tackle the challenging multi-class open set classification problem.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“A particularly fruitful approach seems to be the use of Generative Adversarial Networks (GANs) for detecting out-of-sample cases.[24][25][26]”
From The need for uncertainty quantification in machine-assisted medical decision making (2019) · cites this paper as [26]
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 1×
“Driven by recent research in machine learning, the best performing system can achieve impressive results on both small and large scale multi-class classification tasks [8, 9, 13, 21, 28].”
BackgroundImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155 · cited 2×
“Driven by recent research in machine learning, the best performing system can achieve impressive results on both small and large scale multi-class classification tasks [8, 9, 13, 21, 28].”
BackgroundMethodImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333 · cited 4×
“We trained the generator G on ImageNet12 and made several observations: 1) Several classes blend into unrealistic type of object in terms of colour and global structure.”
BackgroundInfluentialImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380 · cited 1×
“Driven by recent research in machine learning, the best performing system can achieve impressive results on both small and large scale multi-class classification tasks [8, 9, 13, 21, 28].”
BackgroundImage-to-Image Translation with Conditional Adversarial Networks
Isola, P., Zhu, J.-Y., Zhou, T., & Efros, A. A. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.632 · Cited by 22,930 · cited 1×
“GAN has a wide range of applications including, sketch/text to image synthesis [4], image to image translation [10], image editing [19], super resolution [15] and video prediction [6], etc.”
BackgroundAuto-Encoding Variational Bayes
Kingma, D. P., & Welling, M. (2013) · arXiv · doi:10.48550/arxiv.1312.6114 · Cited by 20,067 · cited 2×
“The most prominent are variational autoencoder (VAE) [12] and generative adversarial networks (GAN) [7], among which, GAN is a powerful model of learning arbitrarily complex data distributions.”
BackgroundPhoto-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
Ledig, C., Theis, L., Huszar, F., Caballero, J., Cunningham, A., Acosta, A., Aitken, A., Tejani, A., Totz, J., Wang, Z., et al. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.19 · Cited by 12,696 · cited 1×
“GAN has a wide range of applications including, sketch/text to image synthesis [4], image to image translation [10], image editing [19], super resolution [15] and video prediction [6], etc.”
BackgroundUnsupervised representation learning with deep convolutional generative adversarial networks
Radford, A., Metz, L., & Chintala, S. (2015) · arXiv · doi:10.48550/arxiv.1511.06434 · Cited by 12,235
Improved Techniques for Training GANs
Salimans, T., Goodfellow, I., Zaremba, W., Cheung, V., Radford, A., & Chen, X. (2016) · arXiv · doi:10.48550/arxiv.1606.03498 · Cited by 6,965 · cited 2×
“This is done by using generative adversarial networks (GANs) [7] to generate synthetic samples from unknown classes.”
BackgroundInception-v4, Inception-ResNet and the Impact of Residual Connections on Learning
Szegedy, C., Ioffe, S., Vanhoucke, V., & Alemi, A. (2017) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v31i1.11231 · Cited by 4,636 · cited 1×
“Driven by recent research in machine learning, the best performing system can achieve impressive results on both small and large scale multi-class classification tasks [8, 9, 13, 21, 28].”
Background
Flagged first, then by citations
5 of the 32 references couldn’t be matched to a known paper.
Cited by
Sorted by citations
Items 401–445 of 445
Open-Set Intrusion Detection Based on Adversarial Reciprocal Points Learning and Manifold Mixup
Shen, L. (2025) · International Conference on Algorithms, Software Engineering and Network Security (ASENS) · doi:10.1109/asens64990.2025.11011146
“Ge et al. [5] proposed the generative OpenMax algorithm, which leverages conditional generative adversarial networks to create mixed unknown samples.”
Open-Set Recognition of Environmental Sound Based on KDE-GAN and Attractor–Reciprocal Point Learning
Wu, J., Wang, N., Hong, H., Wang, W., Xing, K., & Jiang, Y. (2025) · Acoustics · doi:10.3390/acoustics7020033 · Cited by 2
“G-OpenMax [22] extends OpenMax by integrating a generative model to create data for unknown classes based on the training data of known classes.”
Energy-based jamming pattern open set recognition via spiking wavelet transformer
Zhong, X., Cui, J., & Ji, Z. (2025) · PLoS ONE · doi:10.1371/journal.pone.0325381 · Cited by 2
Sound: Unknown Object Detection for Autonomous Driving
Chen, D., Ma, X., He, Y., Feng, W., Yang, Q. K., Fu, S., & Tsao, T. (2025) · IEEE International Conference on Mobility, Operations, Services and Technologies (MOST) · doi:10.1109/most65065.2025.00020
“Unlike methods that employ GANs [10], [11] or real auxiliary data [23], [24], we generate negative samples from the training data using the mixup method [25].”
InfluentialTransductive Few-Shot Open-set Learning for HRRPs Classification with Robust Wasserstein Distance
Meng, Y. (2025) · IEEE International Conference on Signal Processing, Communications and Computing (ICSPCC) · doi:10.1109/icspcc66825.2025.11194327
“…in the large-scale setting by augmenting the SoftMax activation to account for the possibility of unseen classes [1], generating artificial outliers [9], [10], improving closed-set accuracy [11], or using placeholders to anticipate novel classes’ distributions with adaptive decision boundaries [2].”
Zhu, L., Hong, L., Chen, Y., et al. (2025) · IEEE International Conference on Signal Processing, Communications and Computing (ICSPCC) · doi:10.1109/icspcc66825.2025.11194649
“Ge et al. [7] proposed G-OpenMax, which uses Generative Adversarial Network (GAN) [8] to generate new-class samples based on the mixed distribution of known classes.”
Towards Effective Open-set Graph Class-incremental Learning
Chen, J., Ma, Z., Fu, S., et al. (2025) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3746027.3754822
“For instance, ODIN [18] boosts unknown detection via temperature scaling and input perturbations, IsoMax [23] separates samples by their distance to isotropic prototypes, and G-OpenMax [10] uses GANs to synthesize unknowns for greater robustness.”
OpenDUN: To Discover Unknown Number of Visual Categories
Wu, S.-C., Ning, M., Dong, W. F., Zheng, Y., Lu, D., & Yuan, L. (2025) · IEEE International Conference on Multimedia and Expo (ICME) · doi:10.1109/icme59968.2025.11209279
“The usage of generative models [5] on the OSR problem has also been explored.”
Xu, D., Hou, J., Song, F., et al. (2025) · IEEE Transactions on Reliability · doi:10.1109/tr.2025.3634616
“Currently, numerous open set recognition (OSR) methods have been proposed in the field of CV. Generative methods, such as G-OpenMax [20], OSR with counterfactual instances (OSRCI) [21], and class-conditional autoencoder (C2AE) [22], acquire new examples with the help of generative adversarial…”
Learning to transport for open set domain generalization
Li, C., Wang, S., Long, Y., et al. (2025) · Pattern Recognition · doi:10.1016/j.patcog.2025.112988
Open Set Oracle Character Recognition via Adaptive Decision Boundary
Huang, S., Liu, Z., Liu, B., Peng, W., & Liu, Y. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-09371-4_16
Evidence Conflict Sampling for Open-set Active Learning
Ning, K.-P., Ke, H.-J., Yao, J., Liu, Y., Tian, Y.-H., & Li, Y. (2026) · International Journal of Computer Vision · doi:10.1007/s11263-025-02600-6 · Cited by 1
Enhancing Open-Set RFF Recognition With cGAN: Generating Multiple Unknown Classes
Sun, H., Zou, C., Wang, Q., et al. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3652146
“Methods such as C2AE [21], OpenGAN [22], and G-Openmax [23] follow this paradigm by integrating generative mechanisms with OSR frameworks to better characterize unknown regions.”
FullAUC Optimization for Open-Set Recognition in Remote Sensing Images
Zhang, X.-L., Huang, Z., Zhang, J., et al. (2026) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2026.3652236
“Ge et al. [39] and Neal et al. [40] used generative adversarial networks to generate counterfactual samples for modeling unknown classes.”
Research on Open-Set Signal Recognition Based on Class Centroids and Comprehensive Decision Strategy
Zhao, L., Li, S., & Jiang, Z. (2025) · IEEE Intl Conf on Dependable, Autonomic and Secure Computing, Intl Conf on Pervasive Intelligence and Computing, Intl Conf on Cloud and Big Data Computing, Intl Conf on Cyber Science and Technology Congress (DASC/PiCom/CBDCom/CyberSciTech) · doi:10.1109/cyberscitech68397.2025.00091
“For example, G-OpenMax employs a conditional Generative Adversarial Network (cGAN) to synthesize unknown class data and predict explicit probability estimates for unknown categories [14].”
Open-Set Recognition of Train Gearbox Faults Based on GlowGAN
Bai, K., Wang, J., Guo, J., et al. (2025) · IEEE International Conference on Sensing, Diagnostics, Prognostics, and Control (SDPC) · doi:10.1109/sdpc68151.2025.11347861
“Ge et al. [3] used generative adversarial networks (GAN) to synthesize approximation samples of a known class, and proposed the G-OpenMax algorithm, which can give explicit probability estimates to the generated approximation samples and retrain the classifiers.”
Yang, R., Fu, Y., Wu, C., & Liu, Z. (2025) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw69036.2025.00547
“[11] extended OpenMax by employing generative adversarial networks (GANs) to synthesize data for novel classes, inspired hybrid architectures combining generation and discrimination [5, 10, 16, 21, 28].”
Liang, Y., Wu, W., Li, H., Cui, L., Ju, J., & Xu, P. (2026) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2026.3667485
“Similarly, Ge et al. [29] integrated synthesized outlier data into OpenMax calibration to improve open-set sensitivity.”
Transforming Closed-Set to Open-Set Diagnosis: Placeholder Learning for Bearing Faults
Wang, C., Yang, T., Kai, Z., et al. (2025) · Global Reliability and Prognostics and Health Management Conference (PHM) · doi:10.1109/phm-xian66756.2025.11427848
“Ge [3] proposed the G-OpenMax method, which extends OpenMax by employing Generative Adversarial Networks (GANs) to synthesize images of novel categories.”
IoT-Oriented Open Set Recognition for Discrete Protocol Messages via a Two-Phase Joint Model
Chen, Y., Li, Q., Chu, R., et al. (2025) · International Conference on Computer and Communications (ICCC) · doi:10.1109/iccc68654.2025.11438052
Open-set hyperspectral image classification via orientational distribution learning
Zhang, H., Jiang, X., & Wang, Y. (2026) · The Egyptian Journal of Remote Sensing and Space Science · doi:10.1016/j.ejrs.2026.03.003
Challenges of Real Life Few-Shot Image Classification
Bennequin, E. (2023) · theses.fr (ABES) · doi:10.70675/5e095960z69ebz4e66za7aczd02b259e695b
A Two‐Stage Intrusion Detection Ensemble Model for Airborne Networks
Liu, W., Zhou, M., Wei, R., et al. (2026) · Transactions on Emerging Telecommunications Technologies · doi:10.1002/ett.70418
FCTL: Feature-level contrastive transfer learning for open set recognition
Zhou, W., Yang, Z., Yang, X., Zhang, J., Lin, L., & Kuruoğlu, E. E. (2026) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2026.104779
Chen, Y., Chen, Q., Liu, T., et al. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3686907
“…our approach can learn more discriminative feature representations at the multimodal feature level; compared with generative methods (such as [45], [50], [51]), we can achieve stable modeling of unknown classes without relying on generated samples or external priors, significantly reducing…”
Sun, L., Xue, R., Zha, H., et al. (2026) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2026.3687604
DOODLE: Diffusion-based Out-of-Distribution Learning for Open-set LiDAR Semantic Segmentation
Oh, C., Kim, H., We, D., et al. (2026) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv61042.2026.00225
“On the other hand, generative model-based methods use au-toencoders and GANs to synthesize or reconstruct samples; e.g. , [17, 40] generate synthetic unknown samples, while [42, 62] detect unknowns by comparing reconstructed and original inputs.”
Minimal discriminative learning for open set recognition
Meng, Y., Ren, H., Liu, L., Wu, M., & Qian, S. (2026) · Multimedia Systems · doi:10.1007/s00530-026-02323-9
“Alongside DMs, generative models (GMs) [14, 31, 51] are explored to determine the boundary between the known and unknown classes by synthesizing mixtures of unknown samples during training.”
InfluentialScattering-Clutter Disentangled Prototypes for SAR Open-Set Recognition
He, X., Cai, Z., Sun, J. F., et al. (2026) · IEEE Geoscience and Remote Sensing Letters · doi:10.1109/lgrs.2026.3691105
“Deep OSR has mainly progressed through discriminative calibration [2], generative pseudo-unknown synthesis [3], [4], and metric learning with feature constraints [5], [6], [7].”
FedPD++: Enhanced Federated Open-Set Recognition with Parameter Disentanglement
Yang, C., Zhu, M., Liu, Y., & Yuan, Y. (2026) · International Journal of Computer Vision · doi:10.1007/s11263-026-02861-9
Yang, F., Yin, Y., Sun, S. Y., et al. (2026) · Measurement Science and Technology · doi:10.1088/1361-6501/ae7fab
Feature space extreme value convergence theorem and copula joint model for open set recognition
Xia, Z., Wu, M., Xiao, F., et al. (2026) · Pattern Recognition · doi:10.1016/j.patcog.2026.114352
Dai, W., Liu, C., Li, B., et al. (2026) · Vehicles · doi:10.3390/vehicles8070155
“Open-set recognition (OSR) frameworks (OpenMax [2], EVT calibration [13], reconstruction-based detection [14], prototype learning [15], and GAN-based synthesis [16]) detect out-of-distribution samples.”
A pretraining-enhanced hyperspherical prototype framework for open-set recognition
Yang, Z., Han, D., Dezert, J., et al. (2026) · Expert Systems with Applications · doi:10.1016/j.eswa.2026.133753
Berti, S., Pasquale, G., & Natale, L. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-31663-9_31
Forging the Unknown: Open-Set Deepfake Attribution via Adaptive Fingerprint Learning
Fang, Y., Han, B., Wang, J., Luo, X., Peng, S., Chen, X., Wen, W., & Cheng, S. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-31663-9_45
Energy-Based Open-Set Active Learning for Object Classification
Lyu, Z., & Beksi, W. J. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-31673-8_38
Wu, H., Liu, J., Wang, Y., et al. (2026) · Drones · doi:10.3390/drones10080615
“Counterfactual generation has been used to synthesise samples near open-set boundaries [31], and Generative OpenMax similarly creates open-set-like samples to refine decision boundaries [32].”
Selective Structural Completion Diffusion Model for HRRP Open-Set Recognition
Wang, A., Li, Y., Chen, W., Shi, L., Feng, Y., Wang, P., Zhang, J., Ning, H., Chen, B., & Liu, H. (2026) · IEEE Transactions on Aerospace and Electronic Systems · doi:10.1109/taes.2026.3725223
An Open-Set Recognition Method for Radar Specific Emitter Identification Based on Open-ResNeSt
Fang, H., Chen, S., Zhang, K., et al. (2026) · Information Communication Technologies Conference (ICTC) · doi:10.1109/ictc70246.2026.11650307
“Comparisons are conducted with state-of-the-art (SOTA) methods, including ARPL, MeDAE, Open-SKResNet, and G-Openmax[19].”
Yi, J., Mu, J., Liu, Z., et al. (2026) · IEEE/CIC International Conference on Communications in China (ICCC) · doi:10.1109/iccc70295.2026.11680410
“Open-set recognition (OSR) was formalized to control open-space risk [4], and deep OSR methods such as OpenMax [5], Generative OpenMax [6], and adversarial reciprocal-point learning [7] extend closed-set classifiers with explicit unknown handling.”
A Method for Open‐Set Recognition of LPI Radar Signal Modulation Based on MGPL
Fang, H., Chen, S., Zhang, K., et al. (2026) · IET Communications · doi:10.1049/cmu2.70205
Zhang, X., Lai, M., Cheng, Y., et al. (2026) · Building and Environment · doi:10.1016/j.buildenv.2026.115287
Han, Z., Zhou, B., Yang, H., Cao, F., & Zhang, Y. (2026) · IEEE Journal of Radio Frequency Identification · doi:10.1109/jrfid.2026.3736073
“Generative methods synthesize or exploit unknown-like data, including the GAN-OpenMax framework [28] and OpenGAN [29].”
Metadata
- Authors
- Ge, Z., Demyanov, S., & Garnavi, R.
- Venue
- Procedings of the British Machine Vision Conference 2017
- Publisher
- British Machine Vision Association
- Published
- 2017
- OpenAlex
- W2963875483
- Cited by
- 472
- References
- 32
- Citations by year
- 2026: 36 · 2025: 59 · 2024: 98 · 2023: 90 · 2022: 63 · 2021: 74 · 2020: 31 · 2019: 12 · 2018: 8 · 2017: 1
- FWCI
- 4.16
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- British Machine Vision Conference (BMVC) (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Keywords
- open-set recognition, OpenMax, generative adversarial networks, image generation, unknown class detection, handwritten digits, handwritten characters
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
