Self-Supervised Visual Feature Learning With Deep Neural Networks: A Survey
Open accessThe DOI matches its Crossref record; 204 of the 206 references it cites were found in a registry or an index.
Abstract
Large-scale labeled data are generally required to train deep neural networks in order to obtain better performance in visual feature learning from images or videos for computer vision applications. To avoid extensive cost of collecting and annotating large-scale datasets, as a subset of unsupervised learning methods, self-supervised learning methods are proposed to learn general image and video features from large-scale unlabeled data without using any human-annotated labels. This paper provides an extensive review of deep learning-based self-supervised general visual feature learning methods from images or videos. First, the motivation, general pipeline, and terminologies of this field are described. Then the common deep neural network architectures that used for self-supervised learning are summarized. Next, the schema and evaluation metrics of self-supervised learning methods are reviewed followed by the commonly used datasets for images, videos, audios, and 3D data, as well as the existing self-supervised visual feature learning methods. Finally, quantitative performance comparisons of the reviewed methods on benchmark datasets are summarized and discussed for both image and video feature learning. At last, this paper is concluded and lists a set of promising future directions for self-supervised visual feature learning.
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
“Jing and Tian [18] summarize the self-supervised feature learning methods on image and video data, and four categories are discussed: generation-based, context-based, free semantic label-based, and cross modal-based, where cross-modal-based methods construct learning task using RGB frame sequence an optical flow sequence, which are unique features in the video.”
From Self-Supervised Learning for Time Series Analysis: Taxonomy, Progress, and Prospects (2024) · cites this paper as [18]
In the citing paper
“Another hot research sub-area is self-supervised learning, which learns discriminative visual representation by solving predefined pretext tasks [194].”
From Empowering things with intelligence: a survey of the progress, challenges, and opportunities in artificial intelligence of things (2021) · cites this paper as [194]
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
Very Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 10×
“Figure is reproduced based on VGG [9].”
BackgroundMethodInfluentialImageNet 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 8×
“Trained with accurate human-annotated labels, the supervised learning methods obtained break-through results on different computer vision applications [1], [4], [8], [16].”
MethodInfluentialHochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318 · cited 1×
“Compared to standard RNN [72], long short term memory (LSTM) uses memory cells to store, modify, and access internal states, to better model the long-term temporal relationships within video frames [73].”
BackgroundFaster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
Ren, S., He, K., Girshick, R., & Sun, J. (2017) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2016.2577031 · Cited by 73,472 · cited 2×
“DUE to the powerful ability to learn different levels of general visual features, deep neural networks have been used as the basic structure to many computer vision applications such as object detection [1], [2], [3], semantic segmentation [4], [5], [6], image captioning [7], etc.”
BackgroundMethodFaster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
Ren, S., He, K., Girshick, R., & Sun, J. (2015) · arXiv · doi:10.48550/arxiv.1506.01497 · Cited by 70,812 · cited 2×
“DUE to the powerful ability to learn different levels of general visual features, deep neural networks have been used as the basic structure to many computer vision applications such as object detection [1], [2], [3], semantic segmentation [4], [5], [6], image captioning [7], etc.”
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 5×
“• ImageNet: The ImageNet dataset [13] contains 1.”
BackgroundInfluentialGradient-based learning applied to document recognition
Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998) · Proceedings of the IEEE · doi:10.1109/5.726791 · Cited by 58,805 · cited 1×
“• MNIST: The MNIST is a dataset of handwritten digits consisting of 70, 000 images while 60, 000 images belong to training set and the rest 10, 000 images are for testing [110].”
BackgroundDistinctive Image Features from Scale-Invariant Keypoints
Lowe, D. G. (2004) · International Journal of Computer Vision · doi:10.1023/b:visi.0000029664.99615.94 · Cited by 55,589
You Only Look Once: Unified, Real-Time Object Detection
Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.91 · Cited by 53,803 · cited 1×
“Recently, many datasets such as MSCOCO [99] and OpenImage [14] have been proposed for object detection and many ConvNet-based models [1], [2], [3], [100], [101], [102], [103], [104] have been proposed and obtained great performance.”
Background
Cited by
Items 501–600 of 1,005
Targeted Self Supervision For Classification On A Small Covid-19 Ct Scan Dataset
Ewen, N., & Khan, N. (2021) · 2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI) · doi:10.1109/isbi48211.2021.9434047 · Cited by 29
“A number of recent works showed that performing self supervised learning on the networks pre-trained using ImageNet prior to performing transfer learning can boost the performance of the classifiers [8][9][10].”
BackgroundMethodInfluentialSelf-supervised prediction of the intention to interact with a service robot
Abbate, G., Giusti, A., Schmuck, V., Çeliktutan, O., & Paolillo, A. (2023) · Robotics and Autonomous Systems · doi:10.1016/j.robot.2023.104568 · Cited by 40
“It is worth noting that in the recent deep learning literature, the term “self-supervised” has a different meaning: it denotes the practice of using pretext tasks [36, 37, 38] for learning useful data representations [39] from large amounts of unlabeled data.”
BackgroundSükei, E., Rumetshofer, E., Schmidinger, N., Mayr, A., Schmidt-Erfurth, U. M., Klambauer, G., & Bogunović, H. (2024) · Scientific Reports · doi:10.1038/s41598-024-78515-y · Cited by 38
Personvit: large-scale self-supervised vision transformer for person re-identification
Hu, B., Wang, X., & Liu, W. (2025) · Machine Vision and Applications · doi:10.1007/s00138-025-01659-y · Cited by 43
“Although research related to person ReID in self-supervised pre-training surfaced relatively late and is therefore scarcer, it showcases tremendous potential due to the avoidance of high-cost annotations in the pre-training phase.”
Lepcha, D. C., Goyal, B., Dogra, A., Alkhayyat, A. H., Sahu, P. K., Ali, A., & Kukreja, V. (2025) · Computer Modeling in Engineering & Sciences · doi:10.32604/cmes.2025.070964 · Cited by 54
PhotoApp: photorealistic appearance editing of head portraits
Mallikarjun, B. R., Tewari, A., Dib, A., Weyrich, T., Bickel, B., Seidel, H., Pfister, H., Matusik, W., Chevallier, L., Elgharib, M. A., et al. (2021) · ACM Transactions on Graphics · doi:10.1145/3450626.3459765 · Cited by 33
Improving Fractal Pre-training
Anderson, C., & Farrell, R. J. (2022) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv51458.2022.00247 · Cited by 41
Wang, Q., Chen, K., Dou, W., & Ma, Y. (2023) · IEEE Transactions on Computational Biology and Bioinformatics · doi:10.1109/tcbb.2023.3246979 · Cited by 29
“However, contrastive learning-based methods are prone to suffer from the dimensional collapse problem [31], especially with the vision Transformer (ViT) architecture.”
BackgroundLiu, S., Luo, X., Fu, K., Wang, M., & Song, Z. (2023) · Frontiers of Computer Science · doi:10.1007/s11704-022-2435-4 · Cited by 29
Contrastive self-supervised learning for diabetic retinopathy early detection
Ouyang, J., Mao, D., Guo, Z., Liu, S., Xu, D. H., & Wang, W. (2023) · Medical & Biological Engineering & Computing · doi:10.1007/s11517-023-02810-5 · Cited by 33
“Inspired by the self-supervised methods presented in [13], this paper proposes a CNN based on contrastive self-supervised learning model (SimCLR-DR) for DR early detection, in which both unlabeled and labeled retina photographs can be used well simultaneously to improve the performance of the model…”
BackgroundMethodMa, J., Lin, W., Tang, X., Zhang, X., Liu, F., & Jiao, L. (2023) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2023.3291357 · Cited by 34
“The common pretext tasks [25] are image jigsaw puzzles, image rotation, and so on.”
BackgroundQCLR: Quantum-LSTM contrastive learning framework for continuous mental health monitoring
Padha, A., & Sahoo, A. (2023) · Expert Systems with Applications · doi:10.1016/j.eswa.2023.121921 · Cited by 38
Lui, C. F., Maged, A., & Xie, M. (2023) · Journal of Intelligent Manufacturing · doi:10.1007/s10845-023-02232-y · Cited by 31
“In reality, however, the available image data are often unlabeled since image labelling is extremely time-consuming and requires expert knowledge from domain experts (Jing & Tian, 2021).”
UAV image object detection based on self-attention guidance and global feature fusion
Bai, J., Hu, H., Liu, X., Zhuang, S., & Wang, Z. (2024) · Image and Vision Computing · doi:10.1016/j.imavis.2024.105262 · Cited by 28
Self-supervised Geometric Perception
Yang, H., Dong, W., Carlone, L., & Koltun, V. (2021) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr46437.2021.01412 · Cited by 41
“Self-supervision has been widely adopted in visual learning [42] to avoid massive human annotation.”
BackgroundSchreyer, M., Sattarov, T., & Borth, D. (2021) · ACM International Conference on AI in Finance (ICAIF) · doi:10.1145/3490354.3494373 · Cited by 33
Deep learning for biosignal control: insights from basic to real-time methods with recommendations
Dillen, A., Steckelmacher, D., Efthymiadis, K., Langlois, K. P., Beir, A. D., Marušič, U., Vanderborght, B., Nowé, A., MEEUSEN, R. R., Ghaffari, F., et al. (2022) · Journal of Neural Engineering · doi:10.1088/1741-2552/ac4f9a · Cited by 30
“Self-supervised learning (Jing and Tian 2019) is a similar approach that is used to learn the relevant structure in EEG data by first learning an unsupervised pretext task, after which the model is further trained on the target task with labeled data (Banville et al 2020, Kostas et al 2021).”
BackgroundMethodA Multi-Lingual Speech Recognition-Based Framework to Human-Drone Interaction
Choutri, K., Lagha, M., Meshoul, S., Batouche, M., Kacel, Y., & Mebarkia, N. (2022) · Electronics · doi:10.3390/electronics11121829 · Cited by 28
“To avoid the burden of further data collection and labelling, a Self-Supervised Learning (SSL) approach, as described in [37], could be used for this purpose.”
MethodWang, Y., Song, D., Wang, W., Rao, S., Wang, X., & Wang, M. (2022) · Neurocomputing · doi:10.1016/j.neucom.2022.09.097 · Cited by 31
COVID-19 detection based on self-supervised transfer learning using chest X-ray images
Li, G., Togo, R., Ogawa, T., & Haseyama, M. (2022) · International Journal of Computer Assisted Radiology and Surgery · doi:10.1007/s11548-022-02813-x · Cited by 32
“Recently, self-supervised learning (SSL) methods have received widespread attention [17].”
BackgroundRemotely Sensed Crop Disease Monitoring by Machine Learning Algorithms: A Review
Zhang, T., Cai, Y., Zhuang, P., & Li, J. (2023) · Unmanned Systems · doi:10.1142/s2301385024500237 · Cited by 28
Nielsen, M., Wenderoth, L., Sentker, T., & Werner, R. (2023) · Bioengineering · doi:10.3390/bioengineering10080895 · Cited by 37
“However, the expectation is that based on the learned generalized and robust input representation, down-stream tasks like image classification can be solved more efficiently, i.e., with a reduced amount of ground truth labels [10].”
BackgroundSpatiotemporal Self-Supervised Learning for Point Clouds in the Wild
Wu, Y., Zhang, T., Ke, W., Susstrunk, S. E., & Salzmann, M. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.00508 · Cited by 38
“While many pre-tasks have been proposed [27], contrastive learning has nowadays become a highly popular choice [30,33,40,41,45].”
BackgroundTaghanaki, S. R., Rainbow, M. J., & Etemad, A. (2023) · IEEE Transactions on Human-Machine Systems · doi:10.1109/thms.2023.3303438 · Cited by 31
“Self-supervised learningisanalternativemethodtofullysupervisedtraining,that avoidstheneedforexpensiveannotationsbyutilizinglarge-scale unlabeleddatatolearngeneralizedrepresentations[35],[36].”
Use of Learning Media to Increase Student Learning Motivation in Junior High Schools
Charline, C., Jo, S., & Frederic, E. (2023) · World Psychology · doi:10.55849/wp.v2i3.402 · Cited by 36
NNG-Mix: Improving Semi-Supervised Anomaly Detection With Pseudo-Anomaly Generation
Dong, H., Frusque, G., Zhao, Y., Chatzi, E., & Fink, O. (2024) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2024.3497801 · Cited by 38
“Self-supervised AD is motivated by the recent success of self-supervised learning [50], where a model learns a generalizable representation from unlabeled data by solving one or several pretext tasks.”
A self-supervised network for image denoising and watermark removal
Tian, C., Xiao, J., Zhang, B., Zuo, W., Zhang, Y., & Lin, C. (2024) · Neural Networks · doi:10.1016/j.neunet.2024.106218 · Cited by 28
Dynamic Spatio-Temporal Graph Reasoning for VideoQA With Self-Supervised Event Recognition
Nie, J., Wang, X. Y., Hou, R., Li, G., Chen, H., & Zhu, W. (2024) · IEEE Transactions on Image Processing · doi:10.1109/tip.2024.3411448 · Cited by 32
“As a subset of unsupervised learning methods, self-supervised learning leverages input data itself as supervisions [22], [33] without relying on expensive human annotations, where the model learns data representations through pre-designed auxiliary tasks and automatically generated pseudo-labels.”
BackgroundJung, J.-W., Shim, H.-J., Kim, J.-H., & Yu, H.-J. (2021) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp39728.2021.9414406 · Cited by 27
“The second architecture resembles the long-term learning of babies, which is also a motivation for self-supervised learning studies [16–18].”
BackgroundSelf-Supervised Multi-Task Pretraining Improves Image Aesthetic Assessment
Pfister, J., Kobs, K., & Hotho, A. (2021) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw53098.2021.00091 · Cited by 33
“Many self-supervised pretext tasks for CNNs [17, 7, 10, 21] teach universally applicable image features using tasks like: rotation classification [22], image part shuffling, or super-resolution [ 17].”
BackgroundPlant phenotyping with limited annotation: Doing more with less
Nagasubramanian, K., Jones, S. E., Singh, A. K., Singh, A., Sarkar, S., & Ganapathysubramanian, B. (2022) · The Plant Phenome Journal · doi:10.1002/ppj2.20051 · Cited by 30
“A recent alternate has been to use large amounts of unlabeled data to pretrain a model, with an approach called self-supervised pretraining (Jing & Tian, 2021; Kar et al., 2021; Misra & van der Maaten, 2020).”
MethodChen, K., Kim, S. J., Gao, Q., & Raschka, S. (2022) · Computational Communication Research · doi:10.5117/ccr2022.1.003.chen · Cited by 28
Saigusa, Y., Sakaino, S., & Tsuji, T. (2022) · IEEE Access · doi:10.1109/access.2022.3185651 · Cited by 32
“In contrast, self-supervised learning has been studied to enhance supervised learning [20], [21].”
BackgroundUnsupervised Continual Learning via Pseudo Labels
He, J., & Zhu, F. M. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-17587-9_2 · Cited by 36
Dominic, J., Bhaskhar, N., Desai, A., Schmidt, A. M., Rubin, E., Gunel, B., Gold, G. E., Hargreaves, B. A., Lenchik, L., Boutin, R. D., et al. (2023) · Bioengineering · doi:10.3390/bioengineering10020207 · Cited by 31
“While most SSL models in computer vision have been used for the downstream task of image classification, segmentation comparatively remains an under-explored task [10].”
Ghelmani, A., & Hammad, A. (2023) · Automation in Construction · doi:10.1016/j.autcon.2023.105001 · Cited by 28
Convolutional neural networks for the classification of glitches in gravitational-wave data streams
Fernandes, T., Vieira, S., Onofre, A., Bustillo, J. C., Torres-Forné, A., & Font, J. A. (2023) · Classical and Quantum Gravity · doi:10.1088/1361-6382/acf26c · Cited by 27
Visual Atoms: Pre-Training Vision Transformers with Sinusoidal Waves
Takashima, S., Hayamizu, R., Inoue, N., Kataoka, H., & Yokota, R. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.01782 · Cited by 40
Predicting stock market trends with self-supervised learning
Ying, Z., Cheng, D., Chen, C., Li, X., Zhu, P. F., Luo, Y., & Liang, Y. (2023) · Neurocomputing · doi:10.1016/j.neucom.2023.127033 · Cited by 27
Self-supervised multi-modality learning for multi-label skin lesion classification
Wang, H., Ahn, E., Bi, L., & Kim, J. (2025) · Computer Methods and Programs in Biomedicine · doi:10.1016/j.cmpb.2025.108729 · Cited by 29
End-to-End Autonomous Driving Without Costly Modularization and 3D Manual Annotation
Guo, M., Zhang, Z., He, Y., Wang, K., Jing, L., & Ling, H. (2025) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2025.3610517 · Cited by 42
Self-citationA survey on learning from graphs with heterophily: recent advances and future directions
Gong, C., Cheng, Y., Yu, J., Xu, C., Shan, C., Luo, S.-Q., & Li, X. (2025) · Frontiers of Computer Science · doi:10.1007/s11704-025-41059-z · Cited by 37
Improving Colonoscopy Lesion Classification Using Semi-Supervised Deep Learning
Golhar, M., Bobrow, T. L., Khoshknab, M. P., Jit, S., Ngamruengphong, S., & Durr, N. J. (2020) · IEEE Access · doi:10.1109/access.2020.3047544 · Cited by 29
“According to [24], there are four common types of proxy tasks: • Generation-based methods: Some part of the data is deliberately removed, and the network is tasked with predicting the missing data.”
Prospective Role of Foundation Models in Advancing Autonomous Vehicles
Wu, J., Gao, B., Gao, J., Yu, J., Chu, H., Yu, Q., Gong, X., Chang, Y. M., Tseng, H. E., Chen, H., et al. (2024) · Research · doi:10.34133/research.0399 · Cited by 37
“In CV, the method of designing pretext tasks according to data attributes includes 4 main categories [21]: generation-based, context-based, free semantic label-based, and cross-modal-based.”
MethodUnsupervised anomaly detection with a temporal continuation, confidence-aware VAE-GAN
Xing, Z., Mehmood, O., & Smith, W. A. P. (2025) · Pattern Recognition · doi:10.1016/j.patcog.2025.111699 · Cited by 28
CNN-Based Acoustic Scene Classification System
Lee, Y., Lim, S., & Kwak, I. (2021) · Electronics · doi:10.3390/electronics10040371 · Cited by 31
“Additionally, we can use self-supervised learning techniques for feature representation learning [34,35] followed by light-weighted traditional machine learning methods.”
MethodData augmentation guided knowledge distillation for environmental sound classification
Tripathi, A. M., & Paul, K. (2022) · Neurocomputing · doi:10.1016/j.neucom.2022.03.025 · Cited by 28
Semisupervised Deep Learning for Image Classification With Distribution Mismatch: A Survey
Calderon-Ramirez, S., Yang, S., & Elizondo, D. A. (2022) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2022.3196326 · Cited by 29
“The process of optimizing a proxy task is also known as self-supervision [44].”
BackgroundPretext Tasks Selection for Multitask Self-Supervised Audio Representation Learning
Zaiem, S., Parcollet, T., Essid, S., & Heba, A. (2022) · IEEE Journal of Selected Topics in Signal Processing · doi:10.1109/jstsp.2022.3195430 · Cited by 31
“SSL recently became a key component to achieve good performance on downstream tasks especially with low-resource setups, either in speech [9], [32], natural language processing [6], [8] or computer vision [3], [4], [33].”
BackgroundSelf-Supervised Learning for Annotation Efficient Biomedical Image Segmentation
Rettenberger, L., Schilling, M. P., Elser, S., Böhland, M., & Reischl, M. (2023) · IEEE Transactions on Biomedical Engineering · doi:10.1109/tbme.2023.3252889 · Cited by 30
“Pseudo annotations are generated by the SSL method to train P applying conventional Supervised Learning (SL) approaches [29], [43].”
MethodFine-Grained Self-Supervision for Generalizable Semantic Segmentation
Zhang, Y., Tian, S., Liao, M., Zhang, Z., Zou, W., & Xu, C. G. (2023) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2023.3285091 · Cited by 26
“Self-supervised methods learn meaningful features by various pretext tasks without human annotation to avoid time-consuming data annotations [51].”
Self-Supervised Dynamic Graph Representation Learning via Temporal Subgraph Contrast
Chen, K., Liu, L., Jiang, L., & Chen, J. (2023) · ACM Transactions on Knowledge Discovery from Data · doi:10.1145/3612931 · Cited by 35
“It has been successfully applied in computer vision [8] and natural language processing [14].”
MethodQu, L., Liu, S., Wang, M., Li, S., Yin, S., & Song, Z. (2023) · Expert Systems with Applications · doi:10.1016/j.eswa.2023.121363 · Cited by 29
Yu, K., Sun, L., Chen, J., Reynolds, M. J., Chaudhary, T., & Batmanghelich, K. N. (2023) · Medical Image Analysis · doi:10.1016/j.media.2023.103062 · Cited by 35
Self-supervised adversarial adaptation network for breast cancer detection
Torabi, M., Rasouli, A., Wu, Q. M. J., Cao, W., & Pourpanah, F. (2024) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2024.108489 · Cited by 26
Training Deep Networks from Zero to Hero: avoiding pitfalls and going beyond
Ponti, M. A., dos Santos, F. P., Ribeiro, L. S. F., & Cavallari, G. B. (2021) · SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI) · doi:10.1109/sibgrapi54419.2021.00011 · Cited by 32
Chen, H., & Martel, A. L. (2025) · Journal of medical imaging · doi:10.1117/1.jmi.12.s2.s22007 · Cited by 28
“SSL trains models to learn generalizable features by solving pretext tasks, such as predicting image rotations or generating contrastive representations, 21 significantly reducing the reliance on labeled medical data.”
Self-Taught Semi-Supervised Anomaly Detection On Upper Limb X-Rays
Spahr, A., Bozorgtabar, B., & Thiran, J. (2021) · IEEE International Symposium on Biomedical Imaging (ISBI) · doi:10.1109/isbi48211.2021.9433771 · Cited by 25
“Learning a representation without using labels is also commonly known as self-supervised learning [24, 25, 26, 27].”
BackgroundZhai, Y., Zhou, W., Sun, B. S., Li, J., Ke, Q., Ying, Z., Gan, J., Mai, C., Labati, R. D., Piuri, V., et al. (2021) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2021.3066195 · Cited by 30
Cross-modal image retrieval with deep mutual information maximization
Gu, C., Bu, J., Zhou, X., Yao, C., Ma, D., Yu, Z., & Yan, X. (2022) · Neurocomputing · doi:10.1016/j.neucom.2022.01.078 · Cited by 30
Self-Supervised Learning for Seismic Image Segmentation From Few-Labeled Samples
Monteiro, B. A. A., Oliveira, H., & dos Santos, J. A. (2022) · IEEE Geoscience and Remote Sensing Letters · doi:10.1109/lgrs.2022.3193567 · Cited by 27
BrainNet: Epileptic Wave Detection from SEEG with Hierarchical Graph Diffusion Learning
Chen, J., Yang, Y., Yu, T., Fan, Y., Mo, X., & Yang, C. (2022) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3534678.3539178 · Cited by 42
Zheng, R., Zhong, Y. Y., Yan, S., Sun, H., Shen, H., & Huang, K. (2022) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2022.3204551 · Cited by 30
“In order to utilize unlabeled data, self-supervised learning (SSL) emerges as a promising approach to learn representations without human annotations [14].”
BackgroundLiao, W., Xiong, H., Wang, Q., Mo, Y., Li, X., Liu, Y., Chen, Z., Huang, S., & Dou, D. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-16452-1_15 · Cited by 35
“To lower the sample size required, the self-supervised learning (SSL) paradigm has been recently proposed to boost the performance of DNN models through learning visual features from images [3] without using labels.”
MethodClaro, P. I. C., Borges, E. P. B. S., Schleder, G. R., Archilha, N. L., Pinto, A. D. S., Carvalho, M. D. O., Driemeier, C., Fazzio, A., & Gouveia, R. F. (2023) · Applied Physics Reviews · doi:10.1063/5.0129324 · Cited by 30
Self-supervised learning for gastritis detection with gastric X-ray images
Li, G., Togo, R., Ogawa, T., & Haseyama, M. (2023) · International Journal of Computer Assisted Radiology and Surgery · doi:10.1007/s11548-023-02891-5 · Cited by 29
“, color and texture) without the manual annotation of labels [8].”
BackgroundAn Overview of the Vision-Based Human Action Recognition Field
Camarena, F., González-Mendoza, M., Chang, L., & Cuevas-Ascencio, R. J. (2023) · Mathematical and Computational Applications · doi:10.3390/mca28020061 · Cited by 32
“Transferring knowledge from one model to another is a standard technique to reduce computational resources and dependency on labeled data [42].”
BackgroundMethodInfluentialSelf-Supervised Nodes-Hyperedges Embedding for Heterogeneous Information Network Learning
Li, M., Zhang, Y., Zhang, W., Chu, Y., Hu, Y., & Yin, B. (2023) · IEEE Transactions on Big Data · doi:10.1109/tbdata.2023.3275374 · Cited by 26
“Over the past five years, self-supervised learning has been widely adopted in graph learning to guide representations based on input data observations [46].”
BackgroundGhaffari, R., Helfroush, M. S., Khosravi, A., Kazemi, K., Danyali, H., & Rutkowski, L. (2023) · Information Fusion · doi:10.1016/j.inffus.2023.101912 · Cited by 25
A Survey on Self-Evolving Autonomous Driving: A Perspective on Data Closed-Loop Technology
Li, X., Wang, Z., Huang, Y., & Chen, H. (2023) · IEEE Transactions on Intelligent Vehicles · doi:10.1109/tiv.2023.3319689 · Cited by 32
“Although manual annotation is not required, SSL requires huge amounts of unannotated data [51].”
BackgroundInfluentialPulver, D., Angkan, P., Hungler, P. C., & Etemad, A. (2023) · INTERNATIONAL CONFERENCE ON MULTIMODAL INTERACTION · doi:10.1145/3577190.3614113 · Cited by 29
“Self-supervised techniques, such as masked autoencoding, have been extensively employed in conjunction with features (rather than raw data) across various domains [24, 38, 41].”
Zhou, C., Zhang, Y., Guo, S. J., Huang, Y., Qiao, X.-N., Wang, R., Zhao, L., Chang, D., Zhao, L.-M., Da, M., et al. (2024) · Frontiers in Oncology · doi:10.3389/fonc.2024.1287995 · Cited by 31
Meta-learners for few-shot weakly-supervised medical image segmentation
Oliveira, H., Gama, P. H. T., Bloch, I., & Junior, R. M. C. (2024) · Pattern Recognition · doi:10.1016/j.patcog.2024.110471 · Cited by 28
A Survey of Trustworthy Representation Learning Across Domains
Zhu, R., Guo, D., Qi, D., Chu, Z., Yu, X., & Li, S. (2024) · ACM Transactions on Knowledge Discovery from Data · doi:10.1145/3657301 · Cited by 31
Cost-effective image recognition of water leakage in metro tunnels using self-supervised learning
Gu, Y., Ai, Q., Xu, Z., Yao, L., Wang, H., Huang, X., & Yuan, Y. (2024) · Automation in Construction · doi:10.1016/j.autcon.2024.105678 · Cited by 26
Hyperspectral anomaly detection with self-supervised anomaly prior
Liu, Y., Jiang, K. Y., Xie, W., Zhang, J., Li, Y., & Fang, L. (2025) · Neural Networks · doi:10.1016/j.neunet.2025.107294 · Cited by 26
Towards Shape Biased Unsupervised Representation Learning for Domain GeneralizationPreprint
Asadi, N., Sarfi, A. M., Hosseinzadeh, M., Karimpour, Z., & Eftekhari, M. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1909.08245 · Cited by 29
“However, in this section, our primary focus is on deep domain adaptation methods due to the significant success of these models[23].”
BackgroundSelf-supervised Incremental Deep Graph Learning for Ethereum Phishing Scam DetectionPreprint
Li, S., Xu, F., Wang, R., & Zhong, S. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.10176 · Cited by 31
“Existing work of SSL[36][25][17] in graph mostly focus on simple graphs, i.e., homogeneous attributed graphs in a relatively small size.”
BackgroundInfluentialTIP: Tabular-Image Pre-training for Multimodal Classification with Incomplete Data
Du, S., Zheng, S., Wang, Y., Bai, W., O’Regan, D. P., & Qin, C. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-72633-0_27 · Cited by 59
A Survey of Data Augmentation in Domain Generalization
Zhong, Y., Zhou, W., & Wang, Z. (2025) · Neural Processing Letters · doi:10.1007/s11063-025-11747-9 · Cited by 32
“The major objective of self-supervised learning [185] is to construct supervised information from the original unsupervised data by designing pretext tasks, which aid the model in learning valuable representations that can be utilized for downstream tasks.”
Ferreira, R. E. P., & Dórea, J. R. R. (2025) · Journal of Dairy Science · doi:10.3168/jds.2024-25650 · Cited by 25
Continual Learning With Neuromorphic Computing: Foundations, Methods, and Emerging Applications
Minhas, M. F., Putra, R. V. W., Awwad, F. R., Hasan, O., & Shafique, M. (2025) · IEEE Access · doi:10.1109/access.2025.3588665 · Cited by 33
“It includes techniques studied in [89], [90], [227]–[235].”
Cognitive Evaluation of Machine Learning Agents
Kadam, S., & Vaidya, V. G. (2020) · Cognitive Systems Research · doi:10.1016/j.cogsys.2020.11.003 · Cited by 28
Grand Challenges in Image Processing
Dufaux, F. (2021) · Frontiers in Signal Processing · doi:10.3389/frsip.2021.675547 · Cited by 28
“More generally, a pretext task can be de fi ned, e.g. image inpainting, colorizing grayscale images, predicting future frames in videos, by withholding some parts of the data and by training the neural network to predict it (Jing and Tian, 2020).”
BackgroundA topological solution to object segmentation and tracking
Tsao, T. R., & Tsao, D. Y. (2022) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2204248119 · Cited by 25
Self-Supervised Joint Learning Fault Diagnosis Method Based on Three-Channel Vibration Images
Zhang, W., Chen, D., & Kong, Y. (2021) · Sensors · doi:10.3390/s21144774 · Cited by 25
“These pseudo-tags do not artificial annotation and can be generated using the image or video attributes; these false tags are then used to learn the characteristics of untagged data [13,16].”
BackgroundTerayama, K., Mizuno, K., Tabeta, S., Sakamoto, S. X., Sugimoto, Y., Sugimoto, K., Fukami, H., Sakagami, M., & Jimenez, L. A. (2021) · Methods in Ecology and Evolution · doi:10.1111/2041-210x.13744 · Cited by 26
DO WE STILL NEED IMAGENET PRE-TRAINING IN REMOTE SENSING SCENE CLASSIFICATION?
Risojević, V., & Stojnić, V. (2022) · The international archives of the photogrammetry, remote sensing and spatial information sciences/International archives of the photogrammetry, remote sensing and spatial information sciences · doi:10.5194/isprs-archives-xliii-b3-2022-1399-2022 · Cited by 30
“Self-supervised learning holds a promise to reduce the need for large labeled datasets in training deep learning models (Jing and Tian, 2020).”
BackgroundWang, J.-Q., Mo, W., Wu, Y., Xu, X.-M., Li, Y., Ye, J., & Lai, X.-B. (2022) · Frontiers in Neuroscience · doi:10.3389/fnins.2022.920820 · Cited by 22
“Deep neural networks are conducive for classification and feature visualization (Alam et al., 2020; Jing and Tian, 2021).”
BackgroundMascolini, A., Cardamone, D., Ponzio, F., Cataldo, S. D., & Ficarra, E. (2022) · BMC Bioinformatics · doi:10.1186/s12859-022-04845-1 · Cited by 24
“This effort includes advances on transfer learning [6], domain adaptation [7], semisupervised learning [8, 9] and self-supervised representation learning [1, 10, 11].”
BackgroundYin, Y., Jin, W., Bai, J., Liu, R. M., & Zhen, H. (2022) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn55064.2022.9892524 · Cited by 24
“In addition, with the increasing development of deep learning, significant progress has been made in the field of self-supervision [17], and many self-supervised networks have been proposed and achieved remarkable results on many computer vision tasks.”
BackgroundSelf Supervised Learning for Few Shot Hyperspectral Image Classification
Braham, N. A. A., Mou, L., Chanussot, J., Mairal, J., & Zhu, X. X. (2022) · IEEE International Geoscience and Remote Sensing Symposium (IGARSS) · doi:10.1109/igarss46834.2022.9884494 · Cited by 24
“In this paper, we propose to leverage recent progress in Self Supervised Learning (SSL) [4] to enable accurate HSI classification with limited labels.”
Method2D Semantic Segmentation: Recent Developments and Future Directions
Guo, Y., Nie, G., Gao, W., & Liao, M. (2023) · Future Internet · doi:10.3390/fi15060205 · Cited by 28
“Weakly supervised learning [67] and unsupervised learning [68] are becoming active research areas.”
BackgroundS 3 Net: Superpixel-Guided Self-Supervised Learning Network for Multitemporal Image Change Detection
Zhan, T., Gong, M., Jiang, X., & Zhang, E. (2023) · IEEE Geoscience and Remote Sensing Letters · doi:10.1109/lgrs.2023.3300308 · Cited by 25
“To deal with this problem, self-supervised learning methods [16], [17] were developed, which are capable of adopting self-defined pseudolabels as supervision for training models to learn useful features of raw data and use these features for subsequent downstream tasks.”
MethodSURVS: A Swin-Unet and game theory-based unsupervised segmentation method for retinal vessel
Wang, T., & Dai, Q. (2023) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2023.107542 · Cited by 25
Seeking an optimal approach for Computer-aided Diagnosis of Pulmonary Embolism
Islam, N. U., Zhou, Z., Gehlot, S., Gotway, M. B., & Liang, J. (2023) · Medical Image Analysis · doi:10.1016/j.media.2023.102988 · Cited by 27
Review of remote sensing change detection in deep learning: Bibliometric and analysis
Yang, B., Mao, Y., Chen, J., Liu, J., Chen, J., Yan, K., & 自然资源部国家卫星海洋应用中心, 北. 1. N. S. O. A. S. (2023) · National Remote Sensing Bulletin · doi:10.11834/jrs.20222156 · Cited by 26
Self-supervised in-domain representation learning for remote sensing image scene classification
Ghanbarzadeh, A., & Soleimani, H. (2024) · Heliyon · doi:10.1016/j.heliyon.2024.e37962 · Cited by 24
Self-supervised deep metric learning for ancient papyrus fragments retrieval
Pirrone, A., Beurton‐Aimar, M., & Journet, N. (2021) · International Journal on Document Analysis and Recognition (IJDAR) · doi:10.1007/s10032-021-00369-1 · Cited by 29
“In recent years, there has been many proposals using this approach for feature learning in a broad sense [26], [32], [37] [16].”
BackgroundSelf-supervised multimodal reconstruction pre-training for retinal computer-aided diagnosis
Hervella, Á. S., Rouco, J., Novo, J., & Ortega, M. (2021) · Expert Systems with Applications · doi:10.1016/j.eswa.2021.115598 · Cited by 28
“The optimization is performed using the Adam optimization algorithm (Kingma & Ba, 2015), with the default decay rates of 𝛽1 = 0.9 and 𝛽2 = 0.999, and batch ize of one image.”
Influential
Metadata
- Authors
- Jing, L., & Tian, Y.
- Affiliations
- The Graduate Center, CUNY (US)
- Venue
- IEEE Transactions on Pattern Analysis and Machine Intelligence
- Published
- 2021-11-01
- Volume
- 43
- Issue
- 11
- Pages
- 4037-4058
- ISSN
- 0162-8828 (ISSN-L)
- 2160-9292
- 1939-3539
- OpenAlex
- W3023371261
- Cited by
- 2,036
- References
- 206
- Citations by year
- 2026: 160 · 2025: 283 · 2024: 354 · 2023: 405 · 2022: 377 · 2021: 305
- FWCI
- 119.51
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE Transactions on Pattern Analysis and Machine Intelligence (publishedVersion)
- PubMed (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- National Science Foundation (IIS-1400802)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Image and Video Retrieval Techniques
- Keywords
- self-supervised learning, visual feature learning, imaging features, video features, deep learning architectures, unlabeled data, audio features, 3D data, benchmark dataset, evaluation metrics
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
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.