Category Anchor-Guided Unsupervised Domain Adaptation for Semantic Segmentation

By Qiming Zhang, Jing Zhang, Dacheng Tao. 2019

312 citations recorded by the source corpus.

Unsupervised domain adaptation (UDA) aims to enhance the generalization capability of a certain model from a source domain to a target domain. UDA is of particular significance since no extra effort is devoted to annotating target domain samples. However, the different data distributions in the two domains, or \emph{domain shift/discrepancy}, inevitably compromise the UDA performance. Although there has been a progress in matching the marginal distributions between two domains, the classifier favors the source domain features and makes incorrect predictions on the target domain due to category-agnostic feature alignment. In this paper, we propose a novel category anchor-guided (CAG) UDA model for semantic segmentation, which explicitly enforces category-aware feature alignment to learn shared discriminative features and classifiers simultaneously. First, the category-wise centroids of the source domain features are used as guided anchors to identify the active features in the target domain and also assign them pseudo-labels. Then, we leverage an anchor-based pixel-level distance loss and a discriminative loss to drive the intra-category features closer and the inter-category features further apart, respectively. Finally, we devise a stagewise training mechanism to reduce the error accumulation and adapt the proposed model progressively. Experiments on both the GTA5$\rightarrow $Cityscapes and SYNTHIA$\rightarrow $Cityscapes scenarios demonstrate the superiority of our CAG-UDA model over the state-of-the-art methods. The code is available at \url{https://github.com/RogerZhangzz/CAG\_UDA}.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

Unsupervised Model Adaptation for Continual Semantic Segmentation (2020, background): “A second class of UDA algorithms directly minimize suitable loss functions that enforce domain alignment (Chen, Gong, and Yang 2017; Wu et al. 2018; Zhang, David, and Gong 2017; Zhang et al. 2019; Lee et al. 2019; Yang and Soatto 2020).”
DACS: Domain Adaptation via Cross-domain Mixed Sampling (2020, background): “This can be targeted at different levels such as at the pixel level [3, 14, 42, 44], feature map level [48, 24, 51, 15, 58, 39, 52] or semantic level [36, 38].”
LoveDA: A Remote Sensing Land-Cover Dataset for Domain Adaptive Semantic Segmentation (2021, methodology): “Recently, the self-training UDA methods have focused on improving the quality of the pseudo-labels [50, 56].”
HRDA: Context-Aware High-Resolution Domain-Adaptive Semantic Segmentation (2022, unknown): “…order to regularize the training and to avoid pseudo-label drift, approaches such as confidence thresholding [48,99], pseudo-label prototypes [41,87,88], and consistency regularization [55, 57, 62] based on data augmentation [1, 10, 49, 51], different context [36,96], domain-mixup…”
MIC: Masked Image Consistency for Context-Enhanced Domain Adaptation (2022, methodology): “The target loss L is defined according to the UDA strategy such as adversarial training [8,21,69,86,87,91] or selftraining [34, 63, 85, 106, 107, 114].”
MetaCorrection: Domain-aware Meta Loss Correction for Unsupervised Domain Adaptation in Semantic Segmentation (2021, unknown): “Our MetaCorrection model, as a self-training based method, also outperforms other related pseudo label learning works [51, 4, 17, 23, 30, 46], demonstrating the effectiveness of the proposed method in alleviating the noise problem.”
CDTrans: Cross-domain Transformer for Unsupervised Domain Adaptation (2021, methodology): “One kind of such UDA methods are based on category-level alignment (Kang et al., 2019; Zhang et al., 2019; Jiang et al., 2020; Li et al., 2021b), which have achieved promising results on public UDA datasets using deep convolution neural networks (CNNs).”
Give Me Your Trained Model: Domain Adaptive Semantic Segmentation without Source Data (2021, methodology): “In addition, we compare our method with previous state-of-the-art works, including source data-dependent methods [2, 4, 23, 42–44, 5] and source data-free data-free methods [20, 19].”
Instance Adaptive Self-Training for Unsupervised Domain Adaptation (2020, methodology): “ng Zhu. arXiv:2008.12197v1 [cs.CV] 27 Aug 2020 2 K. Mei et al. More recently, an alternative research line to reduce domain shift focuses on building schemes based on the self-training (ST) framework [36,2,34,35,20,31]. These works iteratively train the model by using both the labeled source domain ”
Empowering Things With Intelligence: A Survey of the Progress, Challenges, and Opportunities in Artificial Intelligence of Things (2020, methodology): “These issues are related to some well-known machine learning research topics including few-shot learning [31], zero-shot learning [32], meta-learning [33], unsupervised learning [34], semi-supervised learning [35], transfer learning [36], and domain adaptation [37], [38].”