Semi-Supervised and Unsupervised Deep Visual Learning: A Survey

Abstract

State-of-the-art deep learning models are often trained with a large amount of costly labeled training data. However, requiring exhaustive manual annotations may degrade the model's generalizability in the limited-label regime. Semi-supervised learning and unsupervised learning offer promising paradigms to learn from an abundance of unlabeled visual data. Recent progress in these paradigms has indicated the strong benefits of leveraging unlabeled data to improve model generalization and provide better model initialization. In this survey, we review the recent advanced deep learning algorithms on semi-supervised learning (SSL) and unsupervised learning (UL) for visual recognition from a unified perspective. To offer a holistic understanding of the state-of-the-art in these areas, we propose a unified taxonomy. We categorize existing representative SSL and UL with comprehensive and insightful analysis to highlight their design rationales in different learning scenarios and applications in different computer vision tasks. Lastly, we discuss the emerging trends and open challenges in SSL and UL to shed light on future critical research directions.

Degradation Detection in Rice Products via Shape Variations in XCT Simulation-Empowered AI

“…Nevertheless, there is still a need for improvement, especially considering the segmentation of small features as pores or defects in material science, early-stage tumor detection in clinical imaging, or the identification of rotten parts in agricultural products. Deep learning itself is a broad field within the machine learning domain and may also be coarsely classified into two types: supervised and unsupervised methods [29,30]. Supervised methods generally infer knowledge from a training data set having labels associated to it.…”

“…Contrary, unsupervised methods infer knowledge using the training data samples only without any labels. Within segmentation, this means that a segmentation is achieved mostly by clustering, i.e., grouping similar learned features in feature space [29].…”

Advancing systemic disease diagnosis through ophthalmic image‐based artificial intelligence

“…Unlike supervised learning, unsupervised learning does not rely on prelabeled target variables or labels for training, focusing instead on the inherent structure and relationships within the data. 52,53 These algorithms can autonomously uncover hidden patterns or groupings in the data, enabling further exploration of similarities and differences in the information. Currently, due to limited data availability, there has been scant research applying this approach to the analysis of systemic diseases in ophthalmic AI systems.…”

“…As data volumes grow, unsupervised learning could emerge as a more feasible alternative. Unlike supervised learning, unsupervised learning does not rely on prelabeled target variables or labels for training, focusing instead on the inherent structure and relationships within the data 52,53 . These algorithms can autonomously uncover hidden patterns or groupings in the data, enabling further exploration of similarities and differences in the information.…”

Biological representation disentanglement of single-cell data

“…The disentanglement task is a cornerstone problem in machine learning, which has recently attracted much research interest [10][11][12] . This task aims to recover the underlying meaningful factors that shape a complex sample (e.g., a natural image).…”