Abstract:Although semantic segmentation models based on deep neural networks (DNNs) have achieved excellent results, generalizing well from one remote sensing dataset (source domain) to another dataset with different acquisition conditions (target domain) remains a major challenge. Many domain adaptation (DA) approaches have been proposed to address this problem. DA aims to help DNNs learn a generalizable representation space in which source and target domains have similar feature distributions, but most of the existing DA approaches have difficulty in aligning the high-dimensional image representations of two domains directly. In this study, we proposed a model integrating contrastive learning and adversarial learning in a unified framework for aligning two domains in both representation space and spatial layout. Specifically, the model consists of a semantic segmentation network for feature extraction and two branches for DA. The first branch is used for adaptation in representation space directly by a proposed pixelwise contrastive loss, while the second branch is used for adaptation in predicted results to help two domains have similar spatial layouts through a novel but simple entropy-based similarity discriminator. Additionally, a training strategy called category similarity matching sampling was proposed to provide source and target image pairs with similar category composition for each training iteration, which can help the two branches work better. Extensive experiments indicated that the two branches can benefit each other to gain a superior performance and DA pretraining by our methods can achieve impressive results with only a small number of target labeled samples.

Adversarial Domain Adaptation Network for Semantic Role Classification

Multiple Source Domain Adaptation with Adversarial Training of Neural Networks

Adversarial Domain Adaptation with Semantic Consistency for Cross-Domain Image Classification

Domain Adaptation for Remote Sensing Image Semantic Segmentation: An Integrated Approach of Contrastive Learning and Adversarial Learning

Deep Subdomain Adaptation Network for Image Classification

Classification Constrained Discriminator for Domain Adaptive Semantic Segmentation

C2DAN: an Improved Deep Adaptation Network with Domain Confusion and Classifier Adaptation.

Attention-based Adversarial Partial Domain Adaptation

SSF-DAN: Separated Semantic Feature Based Domain Adaptation Network for Semantic Segmentation

Learning Multi-Domain Adversarial Neural Networks for Text Classification.

Domain Adaptation for Semantic Parsing.

Deep Domain Similarity Adaptation Networks for Across Domain Classification

Alleviating the generalization issue in adversarial domain adaptation networks.

Adversarial Domain Adaptation with Domain Mixup

Adversarial Multiple Source Domain Adaptation.

Synthetic Data with Global-Local Level for Adversarial-based Domain Adaptation

Class-conditional domain adaptation for semantic segmentation

Domain Adaptation for Structured Output via Discriminative Patch Representations

CLDA: an Adversarial Unsupervised Domain Adaptation Method with Classifier-Level Adaptation

Multicomponent Adversarial Domain Adaptation: A General Framework.

Align and Adapt: A Two-stage Adaptation Framework for Unsupervised Domain Adaptation