FeatMatch: Feature-Based Augmentation for Semi-supervised Learning

Kuo, C.‐C. Jay; Ma, Chih-Yao; Huang, Jia-Bin; Kira, Zsolt

doi:10.1007/978-3-030-58523-5_28

Cited by 78 publications

(63 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…On PACS, we compare our method with the following baselines and the state-of-the-art semi-supervised learning approaches (i.e., FixMatch [25] and FeatMatch [24]) on the classification accuracy of the target domain using ResNet-18 and ResNet-50. And we also report our performance and baseline results on OfficeHome and miniDomainNet datasets using ResNet-18.…”

Section: B Comparison With Other Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Better Pseudo-label: Joint Domain-aware Label and Dual-classifier for Semi-supervised Domain Generalization

Wang¹,

Liu²,

Shi³

et al. 2021

Preprint

View full text Add to dashboard Cite

With the goal of directly generalizing trained models to unseen target domains, domain generalization (DG), a newly proposed learning paradigm, has attracted considerable attention. Previous DG models usually require a sufficient quantity of annotated samples from observed source domains during training. In this paper, we relax this requirement about full annotation and investigate semi-supervised domain generalization (SSDG) where only one source domain is fully annotated along with the other domains totally unlabeled in the training process. With the challenges of tackling the domain gap between observed source domains and predicting unseen target domains, we propose a novel deep framework via joint domain-aware labels and dualclassifier to produce high-quality pseudo-labels. Concretely, to predict accurate pseudo-labels under domain shift, a domainaware pseudo-labeling module is developed. Also, considering inconsistent goals between generalization and pseudo-labeling: former prevents overfitting on all source domains while latter might overfit the unlabeled source domains for high accuracy, we employ a dual-classifier to independently perform pseudolabeling and domain generalization in the training process. Extensive results on publicly available DG benchmark datasets show the efficacy of our proposed SSDG method compared to the well-designed baselines and the state-of-the-art semi-supervised learning methods.

show abstract

Section: B Comparison With Other Methodsmentioning

confidence: 99%

“…ReMix-Match [23] further improves MixMatch [22]. FeatMatch [24] applies learned feature-based augmentation to consistency loss. FixMatch [25] inherits UDA and ReMixMatch, and combines pseudo-labeling and consistency regularization, finally obtains good performance on SSL benchmarks.…”

Section: B Semi-supervised Learningmentioning

confidence: 99%

Better Pseudo-label: Joint Domain-aware Label and Dual-classifier for Semi-supervised Domain Generalization

Wang¹,

Liu²,

Shi³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Semi-Supervised Learning (SSL) trains models by leveraging both labeled and unlabeled data, and the general idea can be divided into two groups by recent works [43,52,57], i.e., consistency regularization [5,6,13,20,23,30,36,49] and pseudo-labeling [1,3,15,18,42,44]. The first encourages consistent predictions on different augmented versions of the same image, such as virtual adversarial perturbations [30], image mix-up [6,55], grid-masking [8], and even an ensemble of all popular augmentations [5,10,36].…”

Section: Related Workmentioning

confidence: 99%

CrossRectify: Leveraging Disagreement for Semi-supervised Object Detection

Pan¹,

Ye²,

Tang³

et al. 2022

Preprint

View full text Add to dashboard Cite

Semi-supervised object detection (SSOD) has achieved substantial progress in recent years. However, it is observed that the performances of self-labeling SSOD methods remain limited. Based on our experimental analysis, we reveal that the reason behind such phenomenon lies in the mutual error amplification between the pseudo labels and the trained detector. In this study, we propose a Cross Teaching (CT) method, aiming to mitigate the mutual error amplification by introducing the rectification mechanism of pseudo labels. CT simultaneously trains multiple detectors with an identical structure but different parameter initialization. In contrast to existing mutual teaching methods that directly treat predictions from other detectors as pseudo labels, we design the Label Rectification Module (LRM), where the bounding boxes predicted by one detector are rectified using the corresponding boxes predicted by all other detectors with higher confidence scores. In this way, CT can enhance the pseudo label quality compared with self-labeling and existing mutual teaching methods, and reasonably mitigate the mutual error amplification. Over two popular detector structures, i.e., SSD300 and Faster-RCNN-FPN, the proposed CT method obtains consistent improvements and outperforms the state-of-the-art SSOD methods by 2.2% absolute mAP improvements on the Pascal VOC and MS-COCO benchmarks. The code is available at github.com/ machengcheng2016/CrossTeaching-SSOD.

show abstract

“…Semi-supervised learning. Semi-supervised learning for image classification aims in leveraging unlabeled data and improving model performance by making best use of limited labeled data [5,11,30,52]. Three most explored directions are consistency regularization [32,41], entropy minimization [21], and pseudo-labeling [46].…”

Section: Related Workmentioning

confidence: 99%

A Simple Baseline for Low-Budget Active Learning

Pourahmadi¹,

Nooralinejad²,

Pirsiavash³

2021

Preprint

View full text Add to dashboard Cite

Active learning focuses on choosing a subset of unlabeled data to be labeled. However, most such methods assume that a large subset of the data can be annotated. We are interested in low-budget active learning where only a small subset (e.g., 0.2% of ImageNet) can be annotated. Instead of proposing a new query strategy to iteratively sample batches of unlabeled data given an initial pool, we learn rich features by an off-the-shelf self-supervised learning method only once and then study the effectiveness of different sampling strategies given a low budget on a variety of datasets as well as ImageNet dataset. We show that although the state-of-the-art active learning methods work well given a large budget of data labeling, a simple k-means clustering algorithm can outperform them on low budgets. We believe this method can be used as a simple baseline for low-budget active learning on image classification. Code is

show abstract

FeatMatch: Feature-Based Augmentation for Semi-supervised Learning

Cited by 78 publications

References 18 publications

Better Pseudo-label: Joint Domain-aware Label and Dual-classifier for Semi-supervised Domain Generalization

Better Pseudo-label: Joint Domain-aware Label and Dual-classifier for Semi-supervised Domain Generalization

CrossRectify: Leveraging Disagreement for Semi-supervised Object Detection

A Simple Baseline for Low-Budget Active Learning

Contact Info

Product

Resources

About