世上最详细，使用Darknet：Yolov3训练数据

这篇文章格式没调好，把markdown的码放出来了，请看下面链接的文章，是格式正确的：

https://blog.youkuaiyun.com/u012966194/article/details/80004647

 

 

 

我发现有三篇博文写的不错，贴给大家：

https://blog.youkuaiyun.com/lilai619/article/details/79695109

https://blog.youkuaiyun.com/hrsstudy/article/details/65644517?utm_source=itdadao&utm_medium=referral

https://blog.youkuaiyun.com/hrsstudy/article/details/65447947

 

 

 

 

 

 

 

 

 

 

 

 

 

 

*这篇文章将介绍编译darknet框架开始，到整理数据集，到用yolo网络实现一个内部数据集中号码簿的定位。*


# 1. DarkNet的编译
- 当前的环境：亚马逊GPU实例P2：Intel(R) Xeon(R) CPU E5-2686 v4 @ 2.30GHz + K80 + Ubuntu16.04


### 1.1 环境准备：
```bash
sudo -i
#先改软件源：
root@ip-****:~# cd /etc/apt/
root@ip-****:/etc/apt# sudo cp sources.list sources.list.bak
root@ip-****:/etc/apt# sudo vi /etc/apt/sources.list
>>>
  deb http://mirrors.aliyun.com/ubuntu/ xenial main restricted universe multiverse
  deb http://mirrors.aliyun.com/ubuntu/ xenial-security main restricted universe multiverse
  deb http://mirrors.aliyun.com/ubuntu/ xenial-updates main restricted universe multiverse
  deb http://mirrors.aliyun.com/ubuntu/ xenial-backports main restricted universe multiverse
  ##测试版源
  deb http://mirrors.aliyun.com/ubuntu/ xenial-proposed main restricted universe multiverse
  # 源码
  deb-src http://mirrors.aliyun.com/ubuntu/ xenial main restricted universe multiverse
  deb-src http://mirrors.aliyun.com/ubuntu/ xenial-security main restricted universe multiverse
  deb-src http://mirrors.aliyun.com/ubuntu/ xenial-updates main restricted universe multiverse
  deb-src http://mirrors.aliyun.com/ubuntu/ xenial-backports main restricted universe multiverse
  ##测试版源
  deb-src http://mirrors.aliyun.com/ubuntu/ xenial-proposed main restricted universe multiverse
  # Canonical 合作伙伴和附加
  deb http://archive.canonical.com/ubuntu/ xenial partner
  deb http://extras.ubuntu.com/ubuntu/ xenial main


root@ip-****:/etc/apt# sudo apt-get update


#安装rz：
apt install lrzsz


#查看opencv是否可用
pip install --upgrade pip  -i https://pypi.tuna.tsinghua.edu.cn/simple
pip install opencv-python -i https://pypi.tuna.tsinghua.edu.cn/simple
import cv2 #没有报错说明本机opencv可以使用
nvcc -V #查看cuda是否可用


#下载工程并尝试编译
git clone https://github.com/pjreddie/darknet
cd darknet
vim Makefile：
  ... GPU=1
  ... CUDNN=1
  ... OPENCV=1
make -j4


#下载多线程下载器：
wget https://jaist.dl.sourceforge.net/project/kmphpfm/mwget/0.1/mwget_0.1.0.orig.tar.bz2
tar -xjf mwget_0.1.0.orig.tar.bz2
./build
make -j4
make install
```
### 1.2 配置ftp服务器：
因为在远程终端环境，有很多测试图片我们看不到，我们就可以用ftp来从web访问这些图片
```python
from pyftpdlib.authorizers import DummyAuthorizer
from pyftpdlib.handlers import FTPHandler
from pyftpdlib.servers import FTPServer


#实例化虚拟用户，这是FTP验证首要条件<