by madhawav

madhawav / YOLO3-4-Py

A Python wrapper on Darknet. Compatible with YOLO V3.

480 Stars 159 Forks Last release: Not found Apache License 2.0 75 Commits 0 Releases

Available items

No Items, yet!

The developer of this repository has not created any items for sale yet. Need a bug fixed? Help with integration? A different license? Create a request here:


PyPI Badge PyPI Badge Build Status

A Python wrapper on Darknet. Compatible with latest YOLO V3. YOLO 3.0 is a real-time Object Detector by pjreddie.

OutputImage Image source: http://absfreepic.com/free-photos/download/crowded-cars-on-street-4032x2272_48736.html

Google Colab Demo

Refer the following link to preview YOLO3-4-Py in Google Colab: [Google Colab].

Copy the notebook to your drive and run all cells. Ensure that you are in a GPU runtime. You can change the runtime by accessing the menu Runtime/Change runtime type.

What's New?

  • 2020-06-18 - Added a sample Google Colab notebook demonstrating functionality.
  • 2019-01-15 - Added nvidia-docker support.
  • 2018-08-04 - Option to select the preferred GPU -
  • 2018-04-23 - PyPI Release of RC12


1) Python 3.5 2) Python3-Dev (For Ubuntu,

sudo apt-get install python3-dev
) 3) Numpy
pip3 install numpy
4) Cython
pip3 install cython
5) Optionally, OpenCV 3.x with Python bindings. (Tested on OpenCV 3.4.0) - You can use this script to automate Open CV 3.4 installation (Tested on Ubuntu 16.04). - Performance of this approach is better than not using OpenCV. - Installations from PyPI distributions does not use OpenCV.
NOTE: OpenCV 3.4.1 has a bug which causes Darknet to fail. Therefore this wrapper would not work with OpenCV 3.4.1.
More details are available at https://github.com/pjreddie/darknet/issues/502


Installation from PyPI distribution (as described below) is the most convenient approach if you intend to use yolo34py for your projects.

Installation of CPU Only Version

pip3 install yolo34py

Installation of GPU Accelerated Version

pip3 install yolo34py-gpu
NOTE: PyPI Deployments does not use OpenCV due to complexity involved in installation. 
To get best performance, it is recommended to install from source with OpenCV enabled.
NOTE: Make sure CUDA_HOME environment variable is set.

How to run demos in local machine?

1) If you have not installed already, run

python3 setup.py build_ext --inplace
to install library locally. 2) Download "yolov3" model file and config files using
sh download_models.sh
. 3) Run
python3 webcam_demo.py
python3 video_demo.py
python3 image_demo.py

How to run demo using docker?

1) Navigate to docker directory. 2) Copy sample images into the

directory. Or else run input/downloadsampleimages.sh 3) Run
sh run.sh
sh run-gpu.sh
4) Observe the outputs generated in
GPU Version requires nvidia-docker

Installation from Source

1) Set environment variables - To enable GPU acceleration,

export GPU=1
. - To enable OpenCV,
export OPENCV=1

2) Navigate to source root and run

pip3 install .
to install library.

Using a custom version of Darknet

1) Set environment variable DARKNETHOME to download location of darknet. 2) Add DARKNETHOME to LDLIBRARYPATH.

3) Continue instructions for installation from source.

Having trouble?

Kindly raise your issues in the issues section of GitHub repository.

Like to contribute?

Feel free to send PRs or discuss on possible future improvements in issues section. Your contributions are most welcome!

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.