The DOI matches its Crossref record; 113 of the 120 references it cites were found in a registry or an index; none of the 43 sources its claims cite is checked for support yet.
A Survey on Semi-, Self- and Unsupervised Learning for Image Classification
Open accessAbstract
While deep learning strategies achieve outstanding results in computer vision tasks, one issue remains: The current strategies rely heavily on a huge amount of labeled data. In many real-world problems, it is not feasible to create such an amount of labeled training data. Therefore, it is common to incorporate unlabeled data into the training process to reach equal results with fewer labels. Due to a lot of concurrent research, it is difficult to keep track of recent developments. In this survey, we provide an overview of often used ideas and methods in image classification with fewer labels. We compare 34 methods in detail based on their performance and their commonly used ideas rather than a fine-grained taxonomy. In our analysis, we identify three major trends that lead to future research opportunities. 1. State-of-the-art methods are scalable to real-world applications in theory but issues like class imbalance, robustness, or fuzzy labels are not considered. 2. The degree of supervision which is needed to achieve comparable results to the usage of all labels is decreasing and therefore methods need to be extended to settings with a variable number of classes. 3. All methods share some common ideas but we identify clusters of methods that do not share many ideas. We show that combining ideas from different clusters can lead to better performance.
Overview
CoChat Summary
TL;DR
This survey reviews semi-, self-, and unsupervised learning for image classification, comparing 34 methods by their shared ideas and reported performance rather than a fine-grained taxonomy. It identifies three trends: limited real-world scalability, decreasing supervision requirements, and benefits from combining ideas across method clusters.
Key findings
- Across four datasets, reduced supervised methods approach supervised performance: on CIFAR-10, MixMatch and FixMatch exceed 95% accuracy and are roughly 3% below the fully supervised baseline; on CIFAR-100, FixMatch reaches about 77% versus a supervised baseline of about 80%.
- On STL-10, EnAET achieves more than 95%, and FixMatch reports nearly 95% with only 1,000 labels; reduced supervised methods can surpass the supervised baseline by about 20% due to the additional unlabeled data.
- On ILSVRC-2012, SimCLRv2 achieves the best reported Top-1 accuracy of 80.9% and Top-5 accuracy around 96%; only SimCLR, BYOL, and SimCLRv2 are less than 10% worse than the supervised baseline of around 86%.
- The survey concludes that reduced supervised learning reaches comparable results using roughly 10% of labels, and that combining ideas from different clusters can improve performance.
Methods
- Survey design: compares 34 methods for image classification, grouping them by training strategy—one-stage-semi-supervised, one-stage-unsupervised, and multi-stage-semi-supervised—and by common ideas such as cross-entropy, pseudo-labels, pretext tasks, contrastive loss, mutual information, overclustering, and mixup.
- Data and evaluation: analyzes reported results on CIFAR-10, CIFAR-100, STL-10, and ILSVRC-2012, using classification accuracy or cluster accuracy for unsupervised methods.
- Analysis: relies on results reported in the original papers or cross-references rather than recreating every method in a unified setup, and focuses on general trends and specialties instead of small performance differences.
Limitations
- Stated: the comparison uses reported values rather than a unified reimplementation, and differences in architectures, implementations, and adherence to semi-supervised evaluation guidelines may affect comparability.
- Stated: the common-idea framework is fuzzy and incomplete by definition, and the survey excludes related topics such as metric learning, meta learning, GANs, graph networks, pose estimation, segmentation, videos, sketches, and few-/zero-shot learning.
- Evident: the survey does not claim a complete method list and does not present new experiments; its conclusions depend on the availability and consistency of results reported by the reviewed papers.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
0 Valid · 14 Not valid · 29 Not checked · 81 Own results and interpretations
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“For CIFAR-10, CIFAR-100 and ILSVRC-2012 we have a gap of less than 5% left between total supervised and reduced supervised learning.”
§ A Survey on Semi-, Self-and Unsupervised Learning for Image Classification > 2) TREND: HOW MUCH SUPERVISION IS NEEDED?
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The unsupervised methods are separated from the supervised baseline by a clear margin of up to 10%.”
§ A Survey on Semi-, Self-and Unsupervised Learning for Image Classification > TABLE 2.
boosted- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In the comparison, we identified that few common ideas are shared by one-stage-semi-supervised and multi-stagesemi-supervised methods.”
§ A Survey on Semi-, Self-and Unsupervised Learning for Image Classification > 3) TREND: COMBINATION OF COMMON IDEAS
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“For the STL-10 dataset, most methods report a better result than the supervised baseline.”
§ A Survey on Semi-, Self-and Unsupervised Learning for Image Classification > COMPARISON CONCERNING PERFORMANCE
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on 2 assumptions not shown here.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The performance gap between supervised and semi-or self-supervised methods is closing and the number of labels to get comparable results to fully supervised learning is decreasing.”
§ A Survey on Semi-, Self-and Unsupervised Learning for Image Classification > V. CONCLUSION
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
61 more claims
Create a free account to see all 124 claims and every source verdict.
- Probably not a substantial claim. 41
- It may state two separate claims as one. 21
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ A Survey on Semi-, Self-and Unsupervised Learning for Image Classification > COMPARISON CONCERNING USED COMMON IDEAS
“All one-stage-semi-supervised methods use a cross-entropy loss during training”
§ A Survey on Semi-, Self-and Unsupervised Learning for Image Classification > COMPARISON CONCERNING USED COMMON IDEAS
“All one-stage-semi-supervised methods use no CE and often use a pretext task.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ A Survey on Semi-, Self-and Unsupervised Learning for Image Classification > MEAN SQUARED ERROR (MSE)
“= H (P(•|f (x)) + H (P(•|f (x)) | P(•|f (y))) (5”
§ A Survey on Semi-, Self-and Unsupervised Learning for Image Classification > V. CONCLUSION
“MI balances the effects of maximizing the entropy with a uniform distribution for P(•|f (x)) and minimizing the conditional entropy by equalizing P(•|f (x)) and P(•|f (y)).”
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
ImageNet 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
Fully convolutional networks for semantic segmentation
Long, J., Shelhamer, E., & Darrell, T. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298965 · Cited by 38,834
Elements of Information Theory
Cover, T. M., & Thomas, J. A. (2001) · Wiley eBooks · doi:10.1002/0471200611 · Cited by 36,808
Focal Loss for Dense Object Detection
Lin, T.-Y., Goyal, P., Girshick, R., He, K., & Dollar, P. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.324 · Cited by 30,581
He, K., Gkioxari, G., Dollar, P., & Girshick, R. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.322 · Cited by 29,137
(1967) · Cited by 23,719
Learning Multiple Layers of Features from Tiny Images
Krizhevsky, A. (2024) · doi:10.57702/zp44cu3g · Cited by 23,395
Book Review: Pattern Recognition and Machine Learning
Nasrabadi, N. M. (2007) · Journal of Electronic Imaging · doi:10.1117/1.2819119 · Cited by 22,066
On Information and Sufficiency
Kullback, S., & Leibler, R. A. (1951) · The Annals of Mathematical Statistics · doi:10.1214/aoms/1177729694 · Cited by 20,193
Research works printed but not linked: 2
- mey, 2019, Improvability through semi-supervised learning: A survey of theoretical results, arXiv 1908 09574, 1
- 10.24963/ijcai.2019/504 · doi:10.24963/ijcai.2019/504
Cited by
Items 201–247 of 247
Shakya, A. K., Ramola, A., & Vidyarthi, A. (2022) · International Mobile and Embedded Technology Conference (MECON) · doi:10.1109/mecon53876.2022.9752366
Co-Training Method Based on Semi-Decoupling Features for MOOC Learner Behavior Prediction
Wang, H., Xu, L., Huang, Z., & Wang, J. (2022) · Axioms · doi:10.3390/axioms11050223
Unsupervised cross-domain target detection and recognition system based on deep
Gan, B., & Dong, Q. (2021) · International Conference on Artificial Intelligence and Computer Engineering (ICAICE) · doi:10.1109/icaice54393.2021.00040
Semi-supervised Multi-domain Learning for Medical Image Classification
Chavhan, R., Banerjee, B., & Das, N. (2023) · Communications in computer and information science · doi:10.1007/978-3-031-23599-3_3
WLA-Net: A Whole New Light-weight Architecture For Visual Task
Yu, L., Yao, D., Fang, M., & Zhu, L. (2022) · ACM SIGGRAPH International Conference on Virtual-Reality Continuum and its Applications in Industry (VRCAI) · doi:10.1145/3574131.3574458
Can We Teach Functions to an Artificial Intelligence by Just Showing It Enough “Ground Truth”?
Courtois, A., Ehret, T., Arias, P., & Morel, J. (2022) · Lecture notes in mathematics · doi:10.1007/978-3-031-12244-6_31
Image KeyPoints Detector Based On Transformer Of Wavelet Transform
Wu, Z., & Lu, Y. (2023) · International Conference on Intelligent Computing and Signal Processing (ICSP) · doi:10.1109/icsp58490.2023.10248774
Arweiler, J., Ateş, C., Cerquides, J., Koch, R., & Bauer, H.-J. (2023) · Machine Learning and Knowledge Extraction · doi:10.3390/make5040074
Self-citationSharing is Caring: Concurrent Interactive Segmentation and Model Training using a Joint Model
Mikhailov, I. A., Chauveau, B., Bourdel, N., & Bartoli, A. (2023) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw60793.2023.00257
Automatic Clustering by Geometric Method of Tangent Lines with Points
Charandabi, M. M., Khanli, L. M., & Pourbahrami, S. (2023) · International Symposium on Innovative Approaches in Smart Technologies (ISAS) · doi:10.1109/isas60782.2023.10391459
Towards Multi-timescale Online Monitoring of AI Models: Principles and Preliminary Results
Kaakai, F., & Raffi, P.-M. (2023) · HAL
de Castro-Cros, M. (2023) · UPCommons institutional repository (Universitat Politècnica de Catalunya) · doi:10.5821/dissertation-2117-405571
Kokornaczyk, M. O., Acuña, C., y Terán3, A. M., Castelán, M., & Baumgartner, S. (2024) · Research Square · doi:10.21203/rs.3.rs-4020733/v1
Datasets for a data-centric image classification benchmark for noisy and ambiguous label estimation
Schmarje, L., Grossmann, V., Zelenka, C., Dippel, S., Kiko, R., Oszust, M., Pastell, M., Stracke, J., Valros, A., Volkmann, N., et al. (2022) · OpenAgrar · doi:10.5281/zenodo.7152309
Self-citationAn in-depth study of the deep learning pipeline
Courtois, A. (2023) · theses.fr (ABES) · doi:10.70675/309f747bzc879z4b66z8920z51a8b0413b77
Auto-evaluation of skull radiograph accuracy using unsupervised anomaly detection
Watanabe, H., Ezawa, Y., Matsuyama, E., Kondo, Y., Hayashi, N., Maruyama, S., Ogura, T., & Shimosegawa, M. (2024) · Journal of X-Ray Science and Technology · doi:10.3233/xst-230431
Electric Power Load Forecasting Method Based on Contrastive Learning and Attention Mechanism
Bai, Y., Yu, T., Hu, B., Chu, H., & Wang, Q. (2024) · IEEE International Electrical and Energy Conference (CIEEC) · doi:10.1109/cieec60922.2024.10583515
Marefat, A., Kashefi, A., Rastgarpour, M., & Joloudari, J. H. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4907584
On Adversarial Training with Incorrect Labels
Zhao, B. Z. H., Lu, J., Zhou, X., Vatsalan, D., Ikram, M., & Kâafar, M. A. (2024) · Lecture notes in computer science · doi:10.1007/978-981-96-0573-6_9
Multi-path spatial attention for human pose estimation in complex scenes
Ruan, H., & Chen, X. (2024) · International Conference on Computer Information Science and Artificial Intelligence (CISAI) · doi:10.1145/3703187.3703283
Federated Semi-Supervised Medical Image Classification Using Improved Inter-Client Relation Matching
Özdemir, M. S., Sabbagh, M., Erdem, Ç. E., Korçak, Ö., & Uludağ, K. (2025) · International Conference on Data Science and Machine Learning Applications (CDMA) · doi:10.1109/cdma61895.2025.00021
Deep Learning-Based Background Removal using MODNet with Green Spill Correction
Smith, D. G. S. (2025) · INTERANTIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT · doi:10.55041/ijsrem49662
Akbarifar, F., Maghsoodi, N., Dukelow, S. P., Scott, S. H., & Mousavi, P. (2026) · Journal of NeuroEngineering and Rehabilitation · doi:10.1186/s12984-026-02012-8
Peng, L., & Song, Y. J. (2025) · Elsevier eBooks · doi:10.1016/b978-0-443-43950-6.00005-8
Resourceful Learning: Training More Models with Fewer Resources
Robroek, T. T. (. (2024) · IT University Of Copenhagen (IT University of Copenhagen)
Pyo, B. (2024) · DSpace@MIT (Massachusetts Institute of Technology)
Seamless handover for pervasive speech communication
Mitnala, V. N. (2024) · Open Access at Essex (University of Essex)
Shearographic anomaly detection dataset (SADD)
Plassmann, J., Schuler, N., Schuth, M., & von Freymann, G. (2026) · Machine Vision and Applications · doi:10.1007/s00138-026-01948-0
Huo, J., He, F., Lu, C., Feng, W., & Ma, R. (2026) · Analytical Chemistry · doi:10.1021/acs.analchem.4c06621
A Framework for Assessing and Optimising Data Sufficiency in Ultrasound Tongue Imaging
Ani, S. A., Cleland, J., & Zoha, A. (2026) · International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC) · doi:10.5220/0014440800004070
Visual and Textual Common Semantic Spaces for the Analysis of Multimodal Content
Beltrán, L. V. B. (2021) · theses.fr (ABES) · doi:10.70675/a19cbe0dz6c94z4ff6zb70ez024fd8546d17
Towards automatic understanding of narrative audiovisual content
Reboud, A. (2022) · theses.fr (ABES) · doi:10.70675/7b1296fez9f91z4484z8448z8040d28ab833
Soliman, A. (2023) · theses.fr (ABES) · doi:10.70675/32a52416z7918z4954z8b84ze91dda8e76dc
Remote Sensing for Wetland Mapping and Monitoring: Gaps and Future Directions
Singh, R., & Saritha, V. (2026) · Aquatic Conservation Marine and Freshwater Ecosystems · doi:10.1002/aqc.70377
Li, Q., Fan, S., Zhou, L., Wang, Z., & Qi, P. (2026) · Energies · doi:10.3390/en19122809
Visual Analytics Pathway III: Data Engineering
Yousef, I., Shah, S. L., & Gopaluni, R. B. (2026) · Synthesis lectures on emerging engineering technologies · doi:10.1007/978-3-032-22126-1_5
Unsupervised tree species classification with UAV ultra-high resolution multispectral imaging
Lima, L. A. D. S., & Bakuła, K. (2026) · ISPRS annals of the photogrammetry, remote sensing and spatial information sciences · doi:10.5194/isprs-annals-xi-3-2026-351-2026
Using Diffusion Models for Enhancing the Performance of Semi-supervised Image Classifiers
Orăşan, P.-D., Czibula, G., & Albu, A.-I. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-25029-2_23
PRI sequence analysis and anomaly reconstruction method for multi-function radar pulse streams
Binbin, C. H. E. N., Wang, X., Jin, G., Yu, W. A. N. G., & Zhu, D. (2026) · Chinese Journal of Aeronautics · doi:10.1016/j.cja.2026.104399
Diffusion-Based Generative Regularization for Cross-Modal Supervised Learning
Asakura, T., Inoue, N., & Shinoda, K. (2026) · IEEE Access · doi:10.1109/access.2026.3724724
Chao, M., Hu, H., Hui, D., & Zhai, Y. (2026) · REVIEWS ON ADVANCED MATERIALS SCIENCE · doi:10.1515/rams-2025-0298
Next-gen VOC sensing: AI-driven non-invasive detection
Pathak, A. K., Viphavakit, C., & Krishnaswamy, S. (2026) · Elsevier eBooks · doi:10.1016/b978-0-443-29166-1.00009-3
An Empirical Study on the Effects of Missing Labels in Online Continual Learning
Gama, I. A., Alencar, B. M., & Rios, R. A. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-39892-5_20
Artificial intelligence for Alzheimer's disease: From multimodal diagnosis to clinical translation
Pareek, A., Malani, P., & Pareek, A. (2026) · Elsevier eBooks · doi:10.1016/b978-0-443-27722-1.00014-6
Metadata
- Authors
- Schmarje, L., Santarossa, M., Schroder, S.-M., & Koch, R.
- Affiliations
- Christian-Albrechts-Universität zu Kiel (DE)
- Venue
- IEEE Access
- Published
- 2021
- Volume
- 9
- Pages
- 82146-82168
- ISSN
- 2169-3536 (ISSN-L)
- OpenAlex
- W3171229070
- Cited by
- 246
- References
- 120
- Citations by year
- 2026: 24 · 2025: 32 · 2024: 54 · 2023: 51 · 2022: 47 · 2021: 31
- FWCI
- 13.29
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- IEEE Access (publishedVersion, other-oa)
- arXiv (Cornell University) (submittedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- MACAU: Open Access Repository of Kiel University (University Library Kiel) (submittedVersion, cc-by)
- arXiv (Cornell University)
- Funders
- Land Schleswig-Holstein through the Open Access Publikationsfonds Funding Program
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning and Algorithms
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning and Data Classification
- Keywords
- image classification, semi-supervised learning, self-supervised learning, unsupervised learning, limited labeled data, unlabeled data, class imbalance, fuzzy labels, method comparison
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.
A language model drew the claims from the paper on 2026-11-14, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-11-14 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
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
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.
