- Research Article
30
- 10.1016/j.patcog.2021.108436
Source data-free domain adaptation for a faster R-CNN
- Nov 19, 2021
- Pattern Recognition
- Lin Xiong + 4 more +4
Source data-free domain adaptation for a faster R-CNN
Discriminative distribution alignment for domain adaptive object detection
Source data-free domain adaptation for a faster R-CNN
Source data-free domain adaptation for a faster R-CNN
Attention Regularized Laplace Graph for Domain Adaptation.
In leveraging manifold learning in domain adaptation (DA), graph embedding-based DA methods have shown their effectiveness in preserving data manifold through the Laplace graph. However, current graph embedding DA methods suffer from two issues: 1). they are only concerned with preservation of the underlying data structures in the embedding and ignore sub-domain adaptation, which requires taking into account intra-class similarity and inter-class dissimilarity, thereby leading to negative transfer; 2). manifold learning is proposed across different feature/label spaces separately, thereby hindering unified comprehensive manifold learning. In this paper, starting from our previous DGA-DA, we propose a novel DA method, namely A ttention R egularized Laplace G raph-based D omain A daptation (ARG-DA), to remedy the aforementioned issues. Specifically, by weighting the importance across different sub-domain adaptation tasks, we propose the A ttention R egularized Laplace Graph for class aware DA, thereby generating the attention regularized DA. Furthermore, using a specifically designed FEEL strategy, our approach dynamically unifies alignment of the manifold structures across different feature/label spaces, thus leading to comprehensive manifold learning. Comprehensive experiments are carried out to verify the effectiveness of the proposed DA method, which consistently outperforms the state of the art DA methods on 7 standard DA benchmarks, i.e., 37 cross-domain image classification tasks including object, face, and digit images. An in-depth analysis of the proposed DA method is also discussed, including sensitivity, convergence, and robustness.
Read moreSDAN-YOLO:Self-attention domain adaptive network for YOLOv7
Most of the existing unsupervised domain adaptive object detection methods are based on twostage object detectors such as Faster-RCNN because the region proposal network(RPN) module, which onestage detectors don’t have, can help the model achieve instance-level local domain adaption. Although this type of methods have good detection accuracy, the real-time performance cannot meet the requirements of scenarios such as autonomous driving. To this end, we propose a self-attention domain adaptive network suitable for the one-stage detector YOLOv7, which performs the self-attention to extract the position information related to the foreground target as RPN to do instance-level local domain adaptation. At the same time, we also propose a local self-attention augmentation method to suppress the attention noise so that the model can pay attention to the foreground target accurately. Considering the fact that the detection ability of model is from weak to robust during training, we apply coarse-to-fine alignment that global domain adaptation first aligns the the global feature coarsely and transferring to local domain adaptation gradually to finely align the local feature of the foreground target. Our proposed method is validated on Cityspaces, Foggy Cityspaces, KITTI and Sim10K datasets, outperforming existing domain adaptive object detection methods.
Read moreUnsupervised adversarial deep domain adaptation method for potato defects classification
Unsupervised adversarial deep domain adaptation method for potato defects classification
Cross-regional oil palm tree counting and detection via a multi-level attention domain adaptation network
Cross-regional oil palm tree counting and detection via a multi-level attention domain adaptation network
Multilevel Heterogeneous Domain Adaptation Method for Remote Sensing Image Segmentation
Due to more abundant data sources, more various objects of interest, and more time-consuming annotations, there is a large amount of out-of-distribution (OOD) data in the remote sensing field, on which the performance of high-accuracy image segmentation models trained under ideal experimental conditions generally degrades dramatically. Domain adaptation (DA) consequently comes into being, which aims to learn the predictor for the label-scarce target domain of interest with the help of the label-sufficient source domain in the presence of the distribution difference, namely, domain shift, between the two domains. However, the off-the-shelf DA methods for image segmentation not only struggle to cope with the more complex domain shift problems in remote sensing imagery but also almost cannot process heterogeneous data directly without information loss. While the current heterogeneous DA methods mostly still rely on some supervision information from the target domain, which is typically inaccessible in the real world. To overcome these drawbacks, we propose the multilevel heterogeneous unsupervised DA (UDA) method, termed MHDA, which unifies the instance-level DA based on cycle consistency, the feature-level DA based on contrastive learning, and the decision-level DA based on task consistency into a framework to more effectively handle the complex domain shift and heterogeneous data. After that, extensive DA experiments are conducted on the International Society for Photogrammetry and Remote Sensing (ISPRS) dataset, the BigCity dataset constructed by ourselves, and the Wuhan University (WHU) dataset, to explore the effect of each module in MHDA, the necessity of heterogeneous DA, and the effectiveness of multilevel DA. And the results demonstrate that MHDA can achieve superior performance on the remote sensing image segmentation task, compared with several state-of-the-art DA methods.
Read morePCLUDA: A Pseudo-Label Consistency Learning- Based Unsupervised Domain Adaptation Method for Cross-Domain Optical Remote Sensing Image Retrieval
Recent advances in deep learning have dramatically improved the performance of content-based remote sensing image retrieval (CBRSIR) with the same distribution of training set (source domain) and test set (target domain). In fact, their distributions are inconsistent in most cases, which can lead to a dramatic decrease in retrieval performance. Currently, some unsupervised domain adaptation (DA) methods for other remote sensing applications have been proposed to eliminate the inconsistency. However, the current unsupervised DA methods do not make full use of the target domain’s distribution characteristics when delineating its decision boundary. This tends to degrade the cross-domain retrieval performance. In this article, a pseudo-label consistency learning-based unsupervised DA method (PCLUDA) is proposed for cross-domain CBRSIR. Our PCLUDA method minimizes the difference in probability distribution between the target domain and its perturbed output by a pseudo-label self-training and consistency regularization strategy, followed by adjusting the target domain’s decision boundaries to the low-density region. Besides, minimize class confusion (MCC) is introduced to reduce negative transfer caused by large intraclass variance of RSIs. Two cross-domain datasets with 12 cross-domain scenarios are constructed based on six open access datasets to measure DA methods. Experimental results show that our PCLUDA method achieves superior retrieval performances with average retrieval precision improvement by 4.9%–32.3% compared with eight state-of-the-art DA approaches in complex cross-domain scenarios. Furthermore, other experimental results indicate that our PCLUDA can also reach optimal retrieval performances in different kinds of deep learning networks [i.e., vision transformer (ViT) and convolutional neural networks (CNNs)].
Read moreSelf-Training and Adversarial Background Regularization for Unsupervised Domain Adaptive One-Stage Object Detection
Deep learning-based object detectors have shown remarkable improvements. However, supervised learning-based methods perform poorly when the train data and the test data have different distributions. To address the issue, domain adaptation transfers knowledge from the label-sufficient domain (source domain) to the label-scarce domain (target domain). Self-training is one of the powerful ways to achieve domain adaptation since it helps class-wise domain adaptation. Unfortunately, a naive approach that utilizes pseudo-labels as ground-truth degenerates the performance due to incorrect pseudo-labels. In this paper, we introduce a weak self-training (WST) method and adversarial background score regularization (BSR) for domain adaptive one-stage object detection. WST diminishes the adverse effects of inaccurate pseudo-labels to stabilize the learning procedure. BSR helps the network extract discriminative features for target backgrounds to reduce the domain shift. Two components are complementary to each other as BSR enhances discrimination between foregrounds and backgrounds, whereas WST strengthen class-wise discrimination. Experimental results show that our approach effectively improves the performance of the one-stage object detection in unsupervised domain adaptation setting.
Read moreGenerative Adversarial Network-Based Full-Space Domain Adaptation for Land Cover Classification From Multiple-Source Remote Sensing Images
The accuracy of remote sensing image segmentation and classification is known to dramatically decrease when the source and target images are from different sources; while deep learning-based models have boosted performance, they are only effective when trained with a large number of labeled source images that are similar to the target images. In this article, we propose a generative adversarial network (GAN) based domain adaptation for land cover classification using new target remote sensing images that are enormously different from the labeled source images. In GANs, the source and target images are fully aligned in the image space, feature space, and output space domains in two stages via adversarial learning. The source images are translated to the style of the target images, which are then used to train a fully convolutional network (FCN) for semantic segmentation to classify the land cover types of the target images. The domain adaptation and segmentation are integrated to form an end-to-end framework. The experiments that we conducted on a multisource data set covering more than 3500 km <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> with 51 560 256×256 high-resolution satellite images in Wuhan city and a cross-city data set with 11 383 256×256 aerial images in Potsdam and Vaihingen demonstrated that our method exceeded the recent GAN-based domain adaptation methods by at least 6.1% and 4.9% in the mean intersection over union (mIoU) and overall accuracy (OA) indexes, respectively. We also proved that our GAN is a generic framework that can be implemented for other domain transfer methods to boost their performance.
Read moreSemi-supervised domain adaptation for segmentation models on different monitoring settings
Semi-supervised domain adaptation for segmentation models on different monitoring settings
Distribution Alignment and Discriminative Feature Learning for Domain Adaptation in Hyperspectral Image Classification
Domain adaptation (DA) aims to use a well-labeled source domain to predict the labels of the unlabeled or poor-labeled target domain. Most of the existing DA methods focus on the use of feature-level or sample-level information. Recent studies have shown that domain discriminative information is also important for classification. To jointly exploit feature-level information and discriminative information, a new DA method called distribution alignment and discriminative feature learning (DADFL) is proposed for hyperspectral image (HSI) classification in this letter. DADFL incorporates category-discriminative information preservation and structured prediction (SP)-based pseudolabeling into a unified framework to simultaneously reduce distribution and subspace differences between domains. Experimental results on three hyperspectral DA tasks show that the classification performance of the proposed DADFL is better than that of existing DA methods.
Read moreDoes BERT need domain adaptation for clinical negation detection?
Classifying whether concepts in an unstructured clinical text are negated is an important unsolved task. New domain adaptation and transfer learning methods can potentially address this issue. We examine neural unsupervised domain adaptation methods, introducing a novel combination of domain adaptation with transformer-based transfer learning methods to improve negation detection. We also want to better understand the interaction between the widely used bidirectional encoder representations from transformers (BERT) system and domain adaptation methods. We use 4 clinical text datasets that are annotated with negation status. We evaluate a neural unsupervised domain adaptation algorithm and BERT, a transformer-based model that is pretrained on massive general text datasets. We develop an extension to BERT that uses domain adversarial training, a neural domain adaptation method that adds an objective to the negation task, that the classifier should not be able to distinguish between instances from 2 different domains. The domain adaptation methods we describe show positive results, but, on average, the best performance is obtained by plain BERT (without the extension). We provide evidence that the gains from BERT are likely not additive with the gains from domain adaptation. Our results suggest that, at least for the task of clinical negation detection, BERT subsumes domain adaptation, implying that BERT is already learning very general representations of negation phenomena such that fine-tuning even on a specific corpus does not lead to much overfitting. Despite being trained on nonclinical text, the large training sets of models like BERT lead to large gains in performance for the clinical negation detection task.
Read moreDiscriminative and Geometry-Aware Unsupervised Domain Adaptation.
Domain adaptation (DA) aims to generalize a learning model across training and testing data despite the mismatch of their data distributions. In light of a theoretical estimation of the upper error bound, we argue, in this article, that an effective DA method for classification should: 1) search a shared feature subspace where the source and target data are not only aligned in terms of distributions as most state-of-the-art DA methods do but also discriminative in that instances of different classes are well separated and 2) account for the geometric structure of the underlying data manifold when inferring data labels on the target domain. In comparison with a baseline DA method which only cares about data distribution alignment between source and target, we derive three different DA models for classification, namely, close yet discriminative DA (CDDA), geometry-aware DA (GA-DA), and discriminative and GA-DA (DGA-DA), to highlight the contribution of CDDA based on 1), GA-DA based on 2), and, finally, DGA-DA implementing jointly 1) and 2). Using both the synthetic and real data, we show the effectiveness of the proposed approach which consistently outperforms the state-of-the-art DA methods over 49 image classification DA tasks through eight popular benchmarks. We further carry out an in-depth analysis of the proposed DA method in quantifying the contribution of each term of our DA model and provide insights into the proposed DA methods in visualizing both real and synthetic data.
Read moreDomain Adaptive Ship Detection in Optical Remote Sensing Images
With the successful application of the convolutional neural network (CNN), significant progress has been made by CNN-based ship detection methods. However, they often face considerable difficulties when applied to a new domain where the imaging condition changes significantly. Although training with the two domains together can solve this problem to some extent, the large domain shift will lead to sub-optimal feature representations, and thus weaken the generalization ability on both domains. In this paper, a domain adaptive ship detection method is proposed to better detect ships between different domains. Specifically, the proposed method minimizes the domain discrepancies via both image-level adaption and instance-level adaption. In image-level adaption, we use multiple receptive field integration and channel domain attention to enhance the feature’s resistance to scale and environmental changes, respectively. Moreover, a novel boundary regression module is proposed in instance-level adaption to correct the localization deviation of the ship proposals caused by the domain shift. Compared with conventional regression approaches, the proposed boundary regression module is able to make more accurate predictions via the effective extreme point features. The two adaption components are implemented by learning the corresponding domain classifiers respectively in an adversarial training way, thereby obtaining a robust model suitable for both of the two domains. Experiments on both supervised and unsupervised domain adaption scenarios are conducted to verify the effectiveness of the proposed method.
Read moreHealing Sample Selection Bias by Source Classifier Selection
Domain Adaptation (DA) methods are usually carried out by means of simply reducing the marginal distribution differences between the source and target domains, and subsequently using the resultant trained classifier, namely source classifier, for use in the target domain. However, in many cases, the true predictive distributions of the source and target domains can be vastly different especially when their class distributions are skewed, causing the issues of sample selection bias in DA. Hence, DA methods which leverage the source labeled data may suffer from poor generalization in the target domain, resulting in negative transfer. In addition, we observed that many DA methods use either a source classifier or a linear combination of source classifiers with a fixed weighting for predicting the target unlabeled data. Essentially, the labels of the target unlabeled data are spanned by the prediction of these source classifiers. Motivated by these observations, in this paper, we propose to construct many source classifiers of diverse biases and learn the weight for each source classifier by directly minimizing the structural risk defined on the target unlabeled data so as to heal the possible sample selection bias. Since the weights are learned by maximizing the margin of separation between opposite classes on the target unlabeled data, the proposed method is established here as Maximal Margin Target Label Learning (MMTLL), which is in a form of Multiple Kernel Learning problem with many label kernels. Extensive experimental studies of MMTLL against several state-of-the-art methods on the Sentiment and Newsgroups datasets with various imbalanced class settings showed that MMTLL exhibited robust accuracies on all the settings considered and was resilient to negative transfer, in contrast to other counterpart methods which suffered significantly in prediction accuracy.
Read more