Asymmetric Adaptation of Deep Features for Cross-Domain Classification in Remote Sensing Imagery

Ammour, Nassim; Bashmal, Laila; Bazi, Yakoub; Rahhal, Mohamad Mahmoud Al; Zuair, Mansour

doi:10.1109/lgrs.2018.2800642

Cited by 53 publications

(23 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In fact, these differences between source and target domains are quite common on remote sensing images because of different imaging platforms (e.g., satellites and unmanned aerial vehicles) or different imaging sensors (optical sensors, infrared sensors, and SAR sensors). In the past few years, some researchers have explored cross-domain scene classification to enhance the generalization of CNN models and reduce the distribution gap between the target and source domains [162]- [165]. There is much potential for improving domain adaption-based methods for scene classification, such as mapping the feature representations from target and source domains onto a uniform space while preserving the original data structures, designing additional adaptation layers, and optimizing the loss functions.…”

Section: Future Opportunitiesmentioning

confidence: 99%

Remote Sensing Image Scene Classification Meets Deep Learning: Challenges, Methods, Benchmarks, and Opportunities

Cheng

Xie

Han

et al. 2020

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

563

267

View full text Add to dashboard Cite

Remote sensing image scene classification, which aims at labeling remote sensing images with a set of semantic categories based on their contents, has broad applications in a range of fields. Propelled by the powerful feature learning capabilities of deep neural networks, remote sensing image scene classification driven by deep learning has drawn remarkable attention and achieved significant breakthroughs. However, to the best of our knowledge, a comprehensive review of recent achievements regarding deep learning for scene classification of remote sensing images is still lacking. Considering the rapid evolution of this field, this paper provides a systematic survey of deep learning methods for remote sensing image scene classification by covering more than 160 papers. To be specific, we discuss the main challenges of remote sensing image scene classification and survey (1) Autoencoder-based remote sensing image scene classification methods, (2) Convolutional Neural Network-based remote sensing image scene classification methods, and (3) Generative Adversarial Network-based remote sensing image scene classification methods. In addition, we introduce the benchmarks used for remote sensing image scene classification and summarize the performance of more than two dozen of representative algorithms on three commonly-used benchmark data sets. Finally, we discuss the promising opportunities for further research.

show abstract

Section: Future Opportunitiesmentioning

confidence: 99%

Remote Sensing Image Scene Classification Meets Deep Learning: Challenges, Methods, Benchmarks, and Opportunities

Cheng

Xie

Han

et al. 2020

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

563

267

View full text Add to dashboard Cite

show abstract

“…Aimed at enlarging the scale and promoting the efficiency of training datasets, some techniques have been developed [22], such as Transfer Learning [23], Active Learning [24], and others. Ammour et al, used a pretraining network for feature extraction, combined two asymmetric networks for data domain adaptation and classification, mapped the two networks to the same feature space, and carried out post-training for the two networks' weight coefficient adjustment method [25]. Zhou et al carried out migration experiments on data from the same sensor at different times [26].…”

Section: Land-use Classificationmentioning

confidence: 99%

A Combined Convolutional Neural Network for Urban Land-Use Classification with GIS Data

Zeng

et al. 2022

Remote Sensing

View full text Add to dashboard Cite

The classification of urban land-use information has become the underlying database for a variety of applications including urban planning and administration. The lack of datasets and changeable semantics of land-use make deep learning methods suffer from low precision, which prevent improvements in the effectiveness of using AI methods for applications. In this paper, we first used GIS data to produce a well-tagged and high-resolution urban land-use image dataset. Then, we proposed a combined convolutional neural network named DUA-Net for complex and diverse urban land-use classification. The DUA-Net combined U-Net and Densely connected Atrous Spatial Pyramid Pooling (DenseASPP) to extract Remote Sensing Imagers (RSIs) features in parallel. Then, channel attention was used to efficiently fuse the multi-source semantic information from the output of the double-layer network to learn the association between different land-use types. Finally, land-use classification of high-resolution urban RSIs was achieved. Experiments were performed on the dataset of this paper, the publicly available Vaihingen dataset and Potsdam dataset with overall accuracy levels reaching 75.90%, 89.71% and 89.91%, respectively. The results indicated that the complex land-use types with heterogeneous features were more difficult to extract than the single-feature land-cover types. The proposed DUA-Net method proved suitable for high-precision urban land-use classification, which will be of great value for urban planning and national land resource surveying.

show abstract

“…The authors in [19] present an asymmetric adaptation neural network (AANN) technique. Before the variation cycle, they used the features obtained from a pre-trained CNN to feed to a denoising autoencoder to decrease dimensionality.…”

Section: Related Workmentioning

confidence: 99%

SSDAN: Multi-Source Semi-Supervised Domain Adaptation Network for Remote Sensing Scene Classification

et al. 2021

Self Cite

View full text Add to dashboard Cite

We present a new method for multi-source semi-supervised domain adaptation in remote sensing scene classification. The method consists of a pre-trained convolutional neural network (CNN) model, namely EfficientNet-B3, for the extraction of highly discriminative features, followed by a classification module that learns feature prototypes for each class. Then, the classification module computes a cosine distance between feature vectors of target data samples and the feature prototypes. Finally, the proposed method ends with a Softmax activation function that converts the distances into class probabilities. The feature prototypes are also divided by a temperature parameter to normalize and control the classification module. The whole model is trained on both the unlabeled and labeled target samples. It is trained to predict the correct classes utilizing the standard cross-entropy loss computed over the labeled source and target samples. At the same time, the model is trained to learn domain invariant features using another loss function based on entropy computed over the unlabeled target samples. Unlike the standard cross-entropy loss, the new entropy loss function is computed on the model’s predicted probabilities and does not need the true labels. This entropy loss, called minimax loss, needs to be maximized with respect to the classification module to learn features that are domain-invariant (hence removing the data shift), and at the same time, it should be minimized with respect to the CNN feature extractor to learn discriminative features that are clustered around the class prototypes (in other words reducing intra-class variance). To accomplish these maximization and minimization processes at the same time, we use an adversarial training approach, where we alternate between the two processes. The model combines the standard cross-entropy loss and the new minimax entropy loss and optimizes them jointly. The proposed method is tested on four RS scene datasets, namely UC Merced, AID, RESISC45, and PatternNet, using two-source and three-source domain adaptation scenarios. The experimental results demonstrate the strong capability of the proposed method to achieve impressive performance despite using only a few (six in our case) labeled target samples per class. Its performance is already better than several state-of-the-art methods, including RevGrad, ADDA, Siamese-GAN, and MSCN.

show abstract

Asymmetric Adaptation of Deep Features for Cross-Domain Classification in Remote Sensing Imagery

Cited by 53 publications

References 12 publications

Remote Sensing Image Scene Classification Meets Deep Learning: Challenges, Methods, Benchmarks, and Opportunities

Remote Sensing Image Scene Classification Meets Deep Learning: Challenges, Methods, Benchmarks, and Opportunities

A Combined Convolutional Neural Network for Urban Land-Use Classification with GIS Data

SSDAN: Multi-Source Semi-Supervised Domain Adaptation Network for Remote Sensing Scene Classification

Contact Info

Product

Resources

About