Darknet Install and configure Documents

Darknet is a framework designed based on C language for object detection training

Basic configuration

Install baisc tools

$ sudo apt update
$ sudo apt upgrade
$ sudo apt install git wget build-essential python3-dev python3-pip

Install opencv

$ sudo apt install libopencv-dev
$ sudo pip3 install opencv-python

The latest version has been updated to 4.1.x.You can specify the version number to install 3.4.x.

CUDA and cudnn installation (skip without GPU)

Refer thisCUDA and cudnn installation

Install darknet

Download darknet

$ mkdir ~/git
$ cd ~/git
$ git clone https://github.com/pjreddie/darknet.git
$ cd darknet

Cofigure darknet

You need to configure Darknet to fit your PC before completed it.Edit Makefile file

$ vim Makefile

Open GPU

Enable GPU and enable cudnn


edit it,


If you use CPU to train, you should keep it default.

Open opencv

Opencv is required for most object detection in darknet


edit it,



run make with shell,

$ make

wait it to build compiled.

darknet Verify and Use


./darknet imtest data/eagle.jpg

If you look some pictures like this.It mean you install success.

Use yolo model on darknet

Get the trained model file

$ wget https://pjreddie.com/media/files/yolov3.weights

check by pictures

$ ./darknet detect cfg/yolov3.cfg yolov3.weights data/dog.jpg

You would look a picture:


use by webcam

You can use webcam with this comand.

$ ./darknet detector demo cfg/coco.data cfg/yolov3.cfg yolov3.weights

You can see the effect of real-time monitoring.


Not compiled with OpenCV

  1. opencv install fail,unistall the opencv which was install in you PC.And then you can reinstall it fllow this page.
  2. opencv version is invalid,you should check you opencv version >=3.0

nvcc or cudnn related header file not found

cuda and cudnn install fail,reinstall it with this pageInstall cuda and cudnn

GPU error

It is possible that there is not enough memory in GPU, and Darknet requires high GPU.