2020秋招CV面试知识点汇总（持续更新）

此部分记录一下自己在面试中被问到的知识点。（关于算法、编程、语言部分在别的博客中单独记录。）

有小伙伴感兴趣的也可以在评论中留下自己被问到的问题，我看到后会加上。

1、NMS

https://blog.youkuaiyun.com/lanchunhui/article/details/71216463，https://blog.youkuaiyun.com/HappyRocking/article/details/79970627，https://blog.youkuaiyun.com/leviopku/article/details/80886386

2、评价指标recall，PR等

https://www.cnblogs.com/Zhi-Z/p/8728168.html，https://blog.youkuaiyun.com/qq_36653505/article/details/82709281

3、防止过拟合的方法

https://cloud.tencent.com/developer/news/306177

4、BN

https://www.cnblogs.com/eilearn/p/9780696.html

5、pytorch中的数据API

https://blog.youkuaiyun.com/yskyskyer123/article/details/80669894，https://blog.youkuaiyun.com/Mundane_World/article/details/80873074

6、RPN

https://blog.youkuaiyun.com/xys430381_1/article/details/82784645，https://www.cnblogs.com/jiangnanyanyuchen/p/9433791.html，https://www.jianshu.com/p/cbfaa305b887

 7、GPU利用率低的原因

https://blog.youkuaiyun.com/dearcandy/article/details/90715737，https://blog.youkuaiyun.com/qq_32998593/article/details/92849585，https://zhuanlan.zhihu.com/p/53345706

8、线段树

这个知识点真的是第一次被某独角兽公司问到，，，盲点

https://www.cnblogs.com/jason2003/p/9676729.html

https://blog.youkuaiyun.com/huangzihaoal/article/details/81813454

9、Faster RCNN

faster rcnn检测小目标效果不好的原因:https://www.zhihu.com/question/268233942

(注：还被问到了一个开放性的问题：你觉得如何解决这样的小目标检测效果不好的问题？，，emmm自己和面试官的回答与讨论就不写了，大家自行思考吧)

https://www.cnblogs.com/wangguchangqing/p/10384058.html

https://blog.youkuaiyun.com/hust_lmj/article/details/78974348

https://www.julyedu.com/question/big/kp_id/32/ques_id/2103

10、正则化项L1和L2的理解

https://blog.youkuaiyun.com/qq_20412595/article/details/81636105

https://blog.youkuaiyun.com/jinping_shi/article/details/52433975

https://blog.youkuaiyun.com/u014479551/article/details/105452820

11、HRNet系列

https://blog.youkuaiyun.com/weixin_37993251/article/details/88043650

https://segmentfault.com/a/1190000019167646

https://blog.youkuaiyun.com/weixin_40671425/article/details/105600137

12、目标检测之样本不均衡的解决方法

https://zhuanlan.zhihu.com/p/63109691

13、目标检测之one-stage和two-stage网络的区别

https://blog.youkuaiyun.com/weixin_33602281/article/details/86169287?utm_medium=distribute.pc_relevant.none-task-blog-BlogCommendFromMachineLearnPai2-2.channel_param&depth_1-utm_source=distribute.pc_relevant.none-task-blog-BlogCommendFromMachineLearnPai2-2.channel_param

http://www.julyedu.com/question/big/kp_id/32/ques_id/2965

14、1*1的卷积核 有什么用

https://www.zhihu.com/question/56024942

15、梯度消失与梯度爆炸

https://blog.youkuaiyun.com/qq_32002253/article/details/89109214?utm_medium=distribute.pc_relevant.none-task-blog-BlogCommendFromMachineLearnPai2-1.channel_param&depth_1-utm_source=distribute.pc_relevant.none-task-blog-BlogCommendFromMachineLearnPai2-1.channel_param

https://blog.youkuaiyun.com/u011204487/article/details/76383695?utm_medium=distribute.pc_aggpage_search_result.none-task-blog-2~all~first_rank_v2~rank_v25-1-76383695.nonecase

16、目标检测：Feature Pyramid Networks（FPN）

https://blog.youkuaiyun.com/duan19920101/article/details/104238109?utm_medium=distribute.pc_relevant.none-task-blog-baidujs-1&spm=1001.2101.3001.4242

https://blog.youkuaiyun.com/Chunfengyanyulove/article/details/81390796

17、anchor-based 与 anchor-free

https://zhuanlan.zhihu.com/p/62372897

18、小目标检测

https://blog.youkuaiyun.com/ai_study/article/details/101523393

19、优化方法

https://blog.youkuaiyun.com/qq_37098526/article/details/90754356

https://www.cnblogs.com/wuxiangli/p/7258510.html

https://www.cnblogs.com/callyblog/p/8299074.html

20、BN和Dropout在训练和测试时的差别

https://blog.youkuaiyun.com/songyunli1111/article/details/89071021

21、加速卷积计算

https://www.mobibrw.com/2019/17787

22、Focal loss参数及使用场景

https://www.cnblogs.com/king-lps/p/9497836.html

https://www.zhihu.com/question/315802954

23、RoI Pooling 与 RoI Align 的区别

https://zhuanlan.zhihu.com/p/161540817

https://blog.youkuaiyun.com/qq_32172681/article/details/99734978

https://blog.youkuaiyun.com/emiedon/article/details/84727394

24、kmeans及knn的python实现

kmeans：

https://www.cnblogs.com/ahu-lichang/p/7161613.html

https://blog.youkuaiyun.com/qq_30377909/article/details/94596305

knn：

https://www.cnblogs.com/yszd/p/11070192.html

https://www.cnblogs.com/ahu-lichang/p/7151007.html

——————————————————————————————————————————————————————

C++

https://blog.youkuaiyun.com/kuweicai/article/details/82779648

https://www.cnblogs.com/LUO77/p/5771237.html

https://www.cnblogs.com/demian/p/9281966.html

C中new和malloc的区别

https://blog.youkuaiyun.com/nie19940803/article/details/76358673

浅谈C语言字符串结束符'\0'

https://www.cnblogs.com/stefango/p/9329405.html

c++ new delete new[] delete[] 底层实现

https://blog.youkuaiyun.com/cfarmerreally/article/details/54585443

https://blog.youkuaiyun.com/qq_42719751/article/details/90084812

博客资源汇总：

cv岗位 - 深度学习面试知识总结

深度学习岗位面试问题一览

图像处理/CV/ML/DL到HR面总结

图像处理CV/ML/DL 面试题整理（一）、图像处理CV/ML/DL 面试题整理 （二）、图像处理CV/ML/DL 面试题整理 （三）

https://github.com/taizilongxu/interview_python

 

 

——————————————————————————————————————————————————————

概率题

1、一根木棒折成三段，组成三角形的概率

https://www.zybang.com/question/70b34ffdcda26eb97ef1e1586f0d13eb.html

https://www.cnblogs.com/pangxiaodong/archive/2011/08/20/2147283.html

2、100人坐飞机，第一个乘客在座位中随便选一个坐下，第100人正确坐到自己坐位的概率是？

https://www.zhihu.com/question/35950050