![](https://img-blog.csdnimg.cn/20201014180756927.png?x-oss-process=image/resize,m_fixed,h_64,w_64)
域自适应
SHY_VWind
Technical Ants
展开
-
论文解读:Adapting Object Detectors via Selective Cross-Domain Alignment
论文题目:Adapting Object Detectors via Selective Cross-Domain Alignment(CVPR2019)论文主要贡献:以往的域自适应的方法在分类和分割中应用较多,检测中的相关论文比较少,顶会论文大概就4-5篇,而以往的域适应的方法主要针对全局的特征对齐(在loss和距离度量上做文章),这在分类和分割上没什么问题,但是检测任务更多的不是基于全局的...原创 2019-07-29 23:00:40 · 2502 阅读 · 0 评论 -
论文解读:Multi-level Domain Adaptive learning for Cross-Domain Detection
论文题目:Multi-level Domain Adaptive learning for Cross-Domain Detection(Task-CV workshop ICCV2019)论文主要贡献:已有的目标检测自适应方法主要基于全局的特征自适应,没有考虑局部特征自适应(其实是不对的,现有的检测自适应方法有基于全局的,也有局部和实例的特征自适应,分割和分类主要是全局自适应),因此本论文提...原创 2019-07-30 23:32:36 · 1164 阅读 · 0 评论 -
论文解读:Detach and Adapt: Learning Cross-Domain Disentangled Deep Representation
论文题目:Detach and Adapt: Learning Cross-Domain Disentangled Deep Representation(CVPR2018 spotlight)论文思路:主要提出跨域表征分离网络CDRD(Cross-Domain Representation Disentangler (CDRD))以及将CDRD和条件图像翻译相结合的拓展网络ECDRD(Ext...原创 2019-08-04 11:21:45 · 878 阅读 · 0 评论 -
论文解读:Exploring Object Relation in Mean Teacher for Cross-Domain Detection
论文题目:Exploring Object Relation in Mean Teacher for Cross-Domain Detection(CVPR 2019)论文主要贡献:该论文主要也是为了解决跨域的目标检测问题,但是与之前的基于判别的特征方法有很大的不同,主要是利用了半监督学习中SOTA方法Mean Teacher(NIPS 2017:Mean teachers are bette...原创 2019-08-04 13:30:08 · 1579 阅读 · 0 评论 -
论文解读:Conditional Generative Adversarial Network for Structured Domain Adaptation
论文题目:Conditional Generative Adversarial Network for Structured Domain Adaptation论文主要贡献:通过条件生成网络提高分割模型的域自适应能力;论文主要网络结构:基础结构:VGG-19+FCN-8s创新点:在conv1成提取的low-level特征结合noise(通过concat)条件生成出特征和conv...原创 2019-07-28 11:52:43 · 1028 阅读 · 0 评论 -
论文解读:Taking A Closer Look at Domain Shift: Category-level Adversaries for Semantics Consistent
论文题目:Taking A Closer Look at Domain ShiftCategory-level Adversaries for Semantics Consistent Domain Adaptation(CVPR2019)代码地址:https://github.com/RoyalVane/CLAN论文主要贡献:现有的分割自适应方法没有考虑不同类别的对齐差异,有的类别...原创 2019-07-28 15:36:58 · 3736 阅读 · 0 评论