NLNL: Negative Learning for Noisy Labels

Kim, Young Dong; Yim, Junho; Yun, Juseung; Kim, Junmo

doi:10.1109/iccv.2019.00019

Cited by 219 publications

(194 citation statements)

References 53 publications

Supporting

Mentioning

192

Contrasting

Order By: Relevance

“…Kim et al [27] propose the use of the learning method called Negative Learning (NL), where it is used as a complementary label (i.e., a label that is different from the annotation). By using complementary labels, the chances of selecting a true label as a complementary are low, and NL decreases the risk of providing incorrect information.…”

Section: Mixup [55] Is a Technique Proposed For Data Augmentation Thmentioning

confidence: 99%

A Survey on Deep Learning with Noisy Labels: How to train your model when you cannot trust on the annotations?

Cordeiro

Carneiro

2020

2020 33rd SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI)

View full text Add to dashboard Cite

Noisy Labels are commonly present in data sets automatically collected from the internet, mislabeled by nonspecialist annotators, or even specialists in a challenging task, such as in the medical field. Although deep learning models have shown significant improvements in different domains, an open issue is their ability to memorize noisy labels during training, reducing their generalization potential. As deep learning models depend on correctly labeled data sets and label correctness is difficult to guarantee, it is crucial to consider the presence of noisy labels for deep learning training. Several approaches have been proposed in the literature to improve the training of deep learning models in the presence of noisy labels. This paper presents a survey on the main techniques in literature, in which we classify the algorithm in the following groups: robust losses, sample weighting, sample selection, meta-learning, and combined approaches. We also present the commonly used experimental setup, data sets, and results of the state-of-the-art models.

show abstract

Section: Mixup [55] Is a Technique Proposed For Data Augmentation Thmentioning

confidence: 99%

A Survey on Deep Learning with Noisy Labels: How to train your model when you cannot trust on the annotations?

Cordeiro

Carneiro

2020

2020 33rd SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI)

View full text Add to dashboard Cite

show abstract

“…Northcutt et al [14] proposed confident learning for characterizing, identifying, and learning with noisy labels. Kim et al [15] proposed Selective Negative Learning and Positive Learning (SelNLPL) to filter and learn with noisy data. These methods face the problem of discriminating difficulty from mismatched labels.…”

Section: Learning From Noisy Labelsmentioning

confidence: 99%

“…Existing methods based on RGB images with noisy labels usually make a strong assumption that all labels are noisy. These studies mostly work on robust algorithms against noisy labels [13], label cleansing methods finding label errors [14] , or combining them together [15]. It was proven that these classifiers have achieved good accuracy on noisy CIFAR10/100 datasets.…”

Section: Introductionmentioning

confidence: 99%

Remote Sensing Image Scene Classification with Noisy Label Distillation

et al. 2020

View full text Add to dashboard Cite

The widespread applications of remote sensing image scene classification-based Convolutional Neural Networks (CNNs) are severely affected by the lack of large-scale datasets with clean annotations. Data crawled from the Internet or other sources allows for the most rapid expansion of existing datasets at a low-cost. However, directly training on such an expanded dataset can lead to network overfitting to noisy labels. Traditional methods typically divide this noisy dataset into multiple parts. Each part fine-tunes the network separately to improve performance further. These approaches are inefficient and sometimes even hurt performance. To address these problems, this study proposes a novel noisy label distillation method (NLD) based on the end-to-end teacher-student framework. First, unlike general knowledge distillation methods, NLD does not require pre-training on clean or noisy data. Second, NLD effectively distills knowledge from labels across a full range of noise levels for better performance. In addition, NLD can benefit from a fully clean dataset as a model distillation method to improve the student classifier’s performance. NLD is evaluated on three remote sensing image datasets, including UC Merced Land-use, NWPU-RESISC45, AID, in which a variety of noise patterns and noise amounts are injected. Experimental results show that NLD outperforms widely used directly fine-tuning methods and remote sensing pseudo-labeling methods.

show abstract

“…Inspired by the benefit of complementary learning [54], we propose to combine the strength of the three loss functions in Eq. (7), Eq.…”

Section: ) Final Loss Functionmentioning

confidence: 99%

Contrastive Self-Supervised Hashing With Dual Pseudo Agreement

Wang

Zhang

et al. 2020

IEEE Access

View full text Add to dashboard Cite

Recently, unsupervised deep hashing has attracted increasing attention, mainly because of its potential ability to learn binary codes without identity annotations. However, because the labels are predicted by their pretext tasks, unsupervised deep hashing becomes unstable when learning with noisy labels. To mitigate this issue, we propose a simple but effective approach to self-supervised hash learning based on dual pseudo agreement. By adding a consistency constraint, our method can prevent corrupted labels and encourage generalization for effective knowledge distillation. Specifically, we use the refined pseudo labels as a stabilization constraint to train hash codes, which can implicitly encode semantic structures of the data into the learned Hamming space. Based on the stable pseudo labels, we propose a self-supervised hashing method with mutual information and noise contrastive loss. Throughout the process of hash learning, the stable pseudo labels and data distributions collaboratively work together as teachers to guide the binary codes learning process. Extensive experiments on three publicly available datasets demonstrate that the proposed method can consistently outperform state-of-the-art methods by large margins. INDEX TERMS Unsupervised hashing, self-supervised learning, pseudo labels, mutual information.

show abstract

NLNL: Negative Learning for Noisy Labels

Cited by 219 publications

References 53 publications

A Survey on Deep Learning with Noisy Labels: How to train your model when you cannot trust on the annotations?

A Survey on Deep Learning with Noisy Labels: How to train your model when you cannot trust on the annotations?

Remote Sensing Image Scene Classification with Noisy Label Distillation

Contrastive Self-Supervised Hashing With Dual Pseudo Agreement

Contact Info

Product

Resources

About