RDA: Robust Domain Adaptation via Fourier Adversarial Attacking

Article Computer Science, Information Systems

Uncertainty-Aware Unsupervised Domain Adaptation in Object Detection

Dayan Guan et al.

Summary: This paper proposes an uncertainty-aware domain adaptation network (UaDAN) that introduces conditional adversarial learning to align well-aligned and poorly-aligned samples separately. By utilizing an uncertainty metric, UaDAN achieves superior performance through curriculum learning strategy.

IEEE TRANSACTIONS ON MULTIMEDIA (2022)

添加到收藏夹

Article Computer Science, Artificial Intelligence

Scale variance minimization for unsupervised domain adaptation in image segmentation

Dayan Guan et al.

Summary: SVMin aims to improve image segmentation performance in the target domain by minimizing scale variance, maintaining semantic structure consistency within images. This method can be easily integrated with existing UDA techniques, providing consistent performance boost with minimal additional parameters.

PATTERN RECOGNITION (2021)

添加到收藏夹

Proceedings Paper Computer Science, Artificial Intelligence

Unsupervised Domain Adaptive 3D Detection with Multi-Level Consistency

Zhipeng Luo et al.

Summary: Deep learning-based 3D object detection has achieved significant success in the field of autonomous driving, but performance degradation in cross-domain deployment remains a challenge. This paper proposes a novel framework, MLC-Net, for unsupervised 3D domain adaptive detection which outperforms existing methods on standard benchmarks. The key insight is that geometric mismatch is the main factor of domain shift.

2021 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2021) (2021)

添加到收藏夹

Proceedings Paper Computer Science, Artificial Intelligence

Domain Adaptive Video Segmentation via Temporal Consistency Regularization

Dayan Guan et al.

Summary: This paper introduces DA-VSN, a domain adaptive video segmentation network that addresses domain gaps in videos. It consists of cross-domain and intra-domain TCR to achieve temporal consistency for target frames, leading to superior performance compared to multiple baselines in extensive experiments.

2021 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2021) (2021)

添加到收藏夹

Article Computer Science, Information Systems

Part-aware Progressive Unsupervised Domain Adaptation for Person Re-Identification

Fan Yang et al.

Summary: This paper proposes an innovative part-aware progressive adaptation network (PPAN) for UDA-based ReID, which aligns feature distributions from a labeled source domain with those of an unlabeled target domain. A novel progressive adaptation strategy (PAS) is designed to alleviate the negative influence of outlier source identities. Experimental results demonstrate the superior robustness and effectiveness of the proposed unsupervised ReID model on widely-used datasets.

IEEE TRANSACTIONS ON MULTIMEDIA (2021)

添加到收藏夹

Proceedings Paper Computer Science, Artificial Intelligence