Semisupervised Deep Convolutional Neural Networks Using Pseudo Labels for PolSAR Image Classification

Fang, Zheng; Zhang, Gong; Dai, Qijun; Kong, Yingying; Wang, Peng

doi:10.1109/lgrs.2020.3036387

Cited by 13 publications

(5 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, in order to use the capabilities of deep learning to tackle the categorization challenge of PolSAR images with limited data, researchers proposed several semi-supervised methods based on deep learning. Fang et al [28] proposed a semi-supervised 3D-CNN model using pseudo labels. Guo et al [29] combined a memory mechanism and a semi-supervised learning method to construct a semi-supervised method based on a memory convolutional network for PolSAR classification.…”

Section: Introductionmentioning

confidence: 99%

PolSAR Image Classification Based on Relation Network with SWANet

Hua

Zhang

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

Deep learning and convolutional neural networks (CNN) have been widely applied in polarimetric synthetic aperture radar (PolSAR) image classification, and satisfactory results have been obtained. However, there is one crucial issue that still has not been solved. These methods require abundant labeled samples and obtaining the labeled samples of PolSAR images is usually time-consuming and labor-intensive. To obtain better classification results with fewer labeled samples, a new attention-based 3D residual relation network (3D-ARRN) is proposed for PolSAR image. Firstly, a multilayer CNN with residual structure is used to extract depth polarimetric features. Secondly, to extract more important feature information and improve the classification results, a spatial weighted attention network (SWANet) is introduced to concentrate the feature information, which is more favorable for a classification task. Then, the features of training and test samples are integrated and CNN is utilized to compute the score of similarity between training and test samples. Finally, the similarity score is used to determine the category of test samples. Studies on four different PolSAR datasets illustrate that the proposed 3D-ARRN model can achieve higher classification results than other comparison methods with few labeled data.

show abstract

Section: Introductionmentioning

confidence: 99%

PolSAR Image Classification Based on Relation Network with SWANet

Hua

Zhang

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Fang et al. [7] first used a k‐means clustering algorithm to pre‐classify SAR data and obtain corresponding pseudo labels and then used these pseudo labels to train the network. However, this method requires high accuracy of false labels and is prone to receive interference from outliers.…”

Section: Introductionmentioning

confidence: 99%

“…Sun et al [6] adopted a similar training process to improve the generalisation ability of the network by pre-training the weight of the network and then fine-tuning the limited training samples. Fang et al [7] first used a k-means clustering algorithm to pre-classify SAR data and obtain corresponding pseudo labels and then used these pseudo labels to train the network. However, this method requires high accuracy of false labels and is prone to receive interference from outliers.…”

Section: Introductionmentioning

confidence: 99%

Synthetic Aperture Radar image classification based on constrictive learning with limited data

Zhu

2022

IET Radar Sonar & Navi

View full text Add to dashboard Cite

It is difficult to obtain a large number of high-quality labelled Synthetic Aperture Radar (SAR) image data. In order to solve the classification task of an SAR image with limited data, this paper proposes a dual-network classification model of pre-training and fine-tuning based on contrastive learning and pseudo-label training strategy. The algorithm firstly obtains the initial weight of the network through unsupervised pretraining and then takes advantage of the pseudo-label information to realise the fine tuning of the dual network. It effectively reduces the number of samples required during classification. Compared with the classification network without optimisation strategy, the proposed algorithm achieves higher classification accuracy in the ablation study.

show abstract

“…We opt instead to learn an auxiliary model to generate pseudo-labels for the missing labels. The use of pseudolabelling has become popular in unsupervised and semi-supervised learning [11,12,13,14] and has been successful for tasks like image classification [15,16,17,18] and segmentation [19,20,21]. Inspired by these works, we propose a framework for learning a primary and conditional module for (semi-) weakly-supervised dense action anticipation.…”

Section: Introductionmentioning

confidence: 99%

Weakly-Supervised Dense Action Anticipation

Zhang

Chen

Yao

2021

Preprint

View full text Add to dashboard Cite

Dense anticipation aims to forecast future actions and their durations for long horizons. Existing approaches rely on fully-labelled data, i.e. sequences labelled with all future actions and their durations. We present a (semi-) weakly supervised method using only a small number of fullylabelled sequences and predominantly sequences in which only the (one) upcoming action is labelled. To this end, we propose a framework that generates pseudo-labels for future actions and their durations and adaptively refines them through a refinement module. Given only the upcoming action label as input, these pseudo-labels guide action/duration prediction for the future. We further design an attention mechanism to predict context-aware durations. Experiments on the Breakfast and 50Salads benchmarks verify our method's effectiveness; we are competitive even when compared to fully supervised state-of-the-art models. We will make our code available at: https://github.com/zhanghaotong1/WSLVideoDenseAnticipation.

show abstract

Semisupervised Deep Convolutional Neural Networks Using Pseudo Labels for PolSAR Image Classification

Cited by 13 publications

References 19 publications

PolSAR Image Classification Based on Relation Network with SWANet

PolSAR Image Classification Based on Relation Network with SWANet

Synthetic Aperture Radar image classification based on constrictive learning with limited data

Weakly-Supervised Dense Action Anticipation

Contact Info

Product

Resources

About