Object Detection Milestones

本文参考：Object Detection in 20 Years: A Survey.
以下仅对目标检测领域的发展历史做总结性介绍，不描述算法详细细节。

一、 传统算法

Year	Algorithm
\(2001\)	\(Viola \: Jones\: Detectors\)
\(2005\)	\(HOG\: Detector\)
\(2008\)	\(Deformable\: Part-based\: Model (DPM)\)

二. Two-stage Detector

2001-2012是传统算法的时代，从2012年自AlexNet起深度学习开始风起云涌。

Year	Mathod	Algorithm
\(2014\)	\(RCNN\)	\(Selective\: Search+CNN+SVM\)
\(2014\)	\(SPPNet\)	\(CNN+Selective\: Search+SPP+SVM\)
\(2015\)	\(Fast \: RCNN\)	\(CNN +Selective\: Search+ ROI + Softmax\)
\(2015\)	\(Faster \: RCNN\)	\(CNN+RPN+ ROI+BBoxReg+Classify\)

以下列出了每个算法所需要了解的相关算法：
RCNN: selective search, 基于图的图像分割算法Graph-Based Image Segmentation
SPPNet: SPPNet原理
Fast RCNN: ROI Pooling
Faster RCNN: Anchor生成、Region Proposal Network（RPN），NMS非极大值抑制

two-stage中常见的算法

三. One-stage Detector

Year	Method
\(2015\)	\(YOLO\)
\(2016\)	\(SSD\)
\(2017\)	\(FPN\)
\(2017\)	\(RetinaNet\)