Сравнение методов

	Слабо контролируемая классификация изображений	Самообучающаяся классификация изображений
Область	Глубокое обучение	Глубокое обучение
Семейство	Machine learning	Machine learning
Год появления	2014–2016	2018–2020
Автор метода	Multiple contributors; class activation map approach: Zhou et al.	Chen et al. (SimCLR); He et al. (MoCo); Grill et al. (BYOL); Caron et al. (DINO)
Тип	Weakly supervised deep learning paradigm	Pretraining + fine-tuning paradigm
Основополагающий источник	Zhou, B., Khosla, A., Lapedriza, A., Oliva, A., & Torralba, A. (2016). Learning Deep Features for Discriminative Localization. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2921–2929.	Chen, T., Kornblith, S., Norouzi, M., & Hinton, G. (2020). A Simple Framework for Contrastive Learning of Visual Representations. Proceedings of the 37th International Conference on Machine Learning (ICML), PMLR 119, 1597–1607.
Другие названия	WSL image classification, image-level supervised classification, noisy-label image classification, weakly labeled visual recognition	SSL image classification, contrastive visual representation learning, self-supervised visual learning, unsupervised pretraining for image classification
Связанные	5	4
Сводка	Weakly supervised image classification trains convolutional or transformer-based networks using only coarse, incomplete, or noisy supervision — such as image-level category labels, hashtags, or web-scraped tags — without requiring precise bounding boxes or pixel annotations. This dramatically reduces labeling cost while still enabling high-accuracy visual recognition at scale.	Self-supervised image classification trains a deep visual encoder on large unlabeled image datasets by solving proxy tasks — such as predicting which two augmented views of the same image are similar — and then fine-tunes only a lightweight classifier head on labeled examples. Pioneered by frameworks such as SimCLR and MoCo around 2020, it drastically reduces the need for expensive manual annotation while achieving accuracy rivaling fully supervised models.