R-cnn

学习RCNN，SPPNet，Fast RCNN，Faster RCNN，R-FCN。这一系列是个递进关系，也是目标检测使用two-stage方法的一个发展过程。想要更好的理解Faster RCNN和R-FCN，只能把这算法都梳理清楚了，才能明白这个递进过程。它们之间的关系参考博客https://blog.youkuaiyun.com/linolzhang/article/details/54344350

1.选择性搜索对象识别（Selective Search for object recognition）

    首先要明白物体的检测框怎么界定，检测框的标定用了论文《Selective Search for object recognition》的方法，https://blog.youkuaiyun.com/sgfmby1994/article/details/78451748以及https://www.cnblogs.com/zhao441354231/p/5941190.html（这篇解读比较好）中有这篇论文的解读。

2.R-CNN

    论文为《Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation》，论文网址为https://ieeexplore.ieee.org/abstract/document/6909475/metrics#metrics，论文的翻译链接https://blog.youkuaiyun.com/v1_vivian/article/details/78599229。论文提到的边框回归（BBR）问题参考博客https://blog.youkuaiyun.com/v1_vivian/article/details/80292569与博客https://blog.youkuaiyun.com/love1055259415/article/details/80041936

3.SPP-Net

论文为《Spatial Pyramid Pooling in Deep ConvolutionalNetworks for Visual Recognition》，论文网址为https://ieeexplore.ieee.org/abstract/document/7005506/authors#authors，论文解读https://blog.youkuaiyun.com/v1_vivian/article/details/73275259

其他目标检测的方法参考https://blog.youkuaiyun.com/linolzhang/article/details/54344350

4.Fast RCNN

论文为《Fast R-CNN》，论文网址为https://ieeexplore.ieee.org/abstract/document/7410526，论文解读为https://www.cnblogs.com/CZiFan/p/9903518.html

5Faster RCNN

论文为《Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks》论文翻译网址为https://blog.youkuaiyun.com/quincuntial/article/details/79132243，论文的解读https://blog.youkuaiyun.com/i_is_a_energy_man/article/details/78958349（讲的非常基础明白）

转载于:https://www.cnblogs.com/smy123/p/10480120.html