| 摘要: |
| 基于提升去噪锚框的检测Transformer（Detection Transformer with Improved Denoising Anchor Boxes，DINO）是近年来新兴的检测器之一，其凭借卓越的检测性能获得了广泛关注。然而，DINO作为一个监督模型难以从源域数据中学习到通用特征表示，其跨域泛化能力仍然面临巨大挑战，在处理具有领域差异的数据时性能会显著下降。为了提升DINO模型的泛化能力，提出了一种基于特征融合和对齐的领域自适应DINO算法。首先在骨干网络提出多尺度特征融合和全局特征对齐策略，提升模型对目标场景的整体适应性，然后在编码器提出局部特征对齐策略，缓解场景布局等因素对模型跨域性能的影响，最后在解码器提出层次特征融合和目标级对齐策略，提升模型的泛化和判别性能。所提方法在不同跨域任务中将目标领域检测的平均精度（mean Average Precision，mAP）提高了1.5%~17.4%。 |
| 关键词: 目标检测 领域自适应 深度学习 DINO检测器 |
| DOI：10.20079/j.issn.1001-893x.250417001 |
|
| 基金项目:国家自然科学基金青年科学基金（62303433） |
|
| Domain Adaptive DINO with Feature Fusion and Alignment |
| ZHU Sihan |
| (1.Southwest China Institute of Electronic Technology，Chengdu 610036，China;2.National Key Laboratory of Complex Aviation System Simulation，Chengdu 610036，China) |
| Abstract: |
| Detection Transformer with improved denoising anchor boxes(DINO) is one of the emerging detectors in recent years and has gained considerable attention for its outstanding detection performance.However，DINO is difficult to learn generic representations from the source domain samples as a supervised model.Its cross-domain generalization capability still faces significant challenges and it may suffer a performance degradation when dealing with data that has domain shift.To improve the generalization ability of the DINO model， |
| Key words: object detection domain adaptation deep learning DINO detector |
《电讯技术》编辑部 