• 基于fastai的目标检测算法，主要是2018年course
  • ① 2018 coursev2 :  https://nbviewer.org/github/fastai/fastai1/tree/master/courses/ pascal.ipynb   pacal_multi.ipynb   SSD
  • ② jav fastai1.x SSD 没有mAP： https://github/jav0927/course-v3/blob/master/SSD_Object_Detection.ipynb 212
  • ③ mAP计算代码： Calculating mAP (mean average precision) on fastai object detection
    • daveluo的mAP计算非常详细 https://gist.github.com/daveluo/2ab83da32e623864e543d7251e9beef4
      • https://nbviewer.org/gist/daveluo/2ab83da32e623864e543d7251e9beef4
    •  daveluo参考了Sylvain Gugger的代码，Sylvain的mAP代码地址为：https://github.com/sgugger/Deep-Learning/blob/master/mAP/Computing%20the%20mAP%20metric.ipynb
    • fastai 论坛的讨论地址为：http://forums.fast.ai/t/mean-average-precision-map/14345 

---

一、mAP计算过程