1、Inception V1 V2 V3 V4
V1：把google net的某一些大的卷積層換成1*1, 3*3, 5*5的小卷積
V2：提出了Batch Normalization（BN）
V3：提出了分解卷積核
V4：結合ResNet
https://blog.csdn.net/u011021773/article/details/80791650
https://blog.csdn.net/weixin_39953502/article/details/80966046
2、R-CNN系列
R-CNN：https://blog.csdn.net/u014696921/article/details/52824097
Fast-RCNN：https://www.cnblogs.com/skyfsm/p/6806246.html
Faster-RCNN：https://blog.csdn.net/weixin_31866177/article/details/81146722
RoIPoolingh和RoIAlign：https://www.cnblogs.com/wangyong/p/8523814.html
Mask-RCNN：https://blog.csdn.net/jiongnima/article/details/79094159
https://blog.csdn.net/linolzhang/article/details/71774168
3、MobileNet V1 V2
https://blog.csdn.net/mzpmzk/article/details/82976871
https://blog.csdn.net/gbyy42299/article/details/82690105
4、Yolo V1 V2 V3
https://blog.csdn.net/u014380165/article/details/80202337
https://blog.csdn.net/zfq740695564/article/details/79754578
5、NMS和Soft-NMS
https://www.cnblogs.com/zf-blog/p/8532228.html
https://blog.csdn.net/u014380165/article/details/79502197
6、交叉熵
https://blog.csdn.net/tsyccnh/article/details/79163834