[论文解读] Unsupervised Domain Adaptation for Object Detection via Cross-Domain Semi-Supervised Learning

Fuxun Yu, Di Wang+8 more|arXiv (Cornell University)|Nov 17, 2019

Domain Adaptation and Few-Shot Learning参考文献 35被引用 36

一句话总结

CDSSL 通过 (1) 将源数据转移到面向检测的中间域并 (2) 进行带有不平衡采样与基于置信度的标签锐化的迭代自训练，消除域间差距，在多个域适配场景中达到最先进的结果。

ABSTRACT

Current state-of-the-art object detectors can have significant performance drop when deployed in the wild due to domain gaps with training data. Unsupervised Domain Adaptation (UDA) is a promising approach to adapt models for new domains/environments without any expensive label cost. However, without ground truth labels, most prior works on UDA for object detection tasks can only perform coarse image-level and/or feature-level adaptation by using adversarial learning methods. In this work, we show that such adversarial-based methods can only reduce the domain style gap, but cannot address the domain content distribution gap that is shown to be important for object detectors. To overcome this limitation, we propose the Cross-Domain Semi-Supervised Learning (CDSSL) framework by leveraging high-quality pseudo labels to learn better representations from the target domain directly. To enable SSL for cross-domain object detection, we propose fine-grained domain transfer, progressive-confidence-based label sharpening and imbalanced sampling strategy to address two challenges: (i) non-identical distribution between source and target domain data, (ii) error amplification/accumulation due to noisy pseudo labeling on the target domain. Experiment results show that our proposed approach consistently achieves new state-of-the-art performance (2.2% - 9.5% better than prior best work on mAP) under various domain gap scenarios. The code will be released.

研究动机与目标

• 在没有目标标签的域移位下，通过利用自监督学习启发的伪标签来推动鲁棒的目标检测。
• 解决超越粗粒度特征对齐的样式与内容分布差异（对象密度、上下文）的问题。
• 提出一个两阶段框架（域转移 + 迭代自训练）以提升伪标签质量。
• 在合成到真实、跨摄像头以及普通到雾天等域适配场景下评估 CDSSL，并与以往的 SOTA 进行比较。

提出的方法

• 通过检测导向的细粒度 CycleGAN，将域转移到一个中间域，使用感受野受限以在保持前景/内容的同时匹配目标域风格。
• 在中间域上训练初始伪标签标注器，以在目标域上生成高质量的伪标签。
• 通过将标注的源/中间数据与伪标注的目标数据结合，执行迭代的半监督自训练。
• 通过不均衡的小批量采样来减缓伪标签误差累积：对源/中间数据进行过采样、对目标伪标记数据进行欠采样。
• 应用基于置信度的硬标签，并在各轮中逐步提高置信阈值，以避免强化错误的伪标签。

实验结果

研究问题

• RQ1 域风格转移是否足以降低分布差距，从而使基于 SSL 的自训练在检测任务中有效？
• RQ2 在无监督域适配下，带有偏向性（不平衡）采样与渐进式标签锐化的迭代自训练是否提升目标域检测性能？
• RQ3 CDSSL 的各组件（域转移、不平衡采样、标签锐化）如何在多样化域移位中提升检测性能？

主要发现

• CDSSL 在 Sim2City、KITTI2City 和 City2Foggy 的适配基准上，始终以 2.2%–9.5% mAP 超越之前的最先进水平。
• 先进行中间域的域转移再进行自训练，所得提升大于仅自训练或仅域转移。
• 不均衡小批量采样和渐进式基于置信度的标注可减缓来自嘈杂伪标签的错误累积并稳定训练。
• 细粒度、面向检测的域转移（更小的贴片尺寸）比粗粒度的全局翻译带来更大收益。
• 在多类别适配中，CDSSL 在若干类别上缩小域差距，并在域转移后对初始表现较弱的类别显示显著提升。