Darknet/YOLO

Browse Hank.ai's announcement in May 2023 on officially sponsoring the Darknet/YOLO codebase.

Download Darknet/YOLO

• Download the latest version of Darknet/YOLO from here: https://github.com/hank-ai/darknet
• See the instructions on how to build Darknet/YOLO
• If you'd like additional assistance, join the Darknet/YOLO Discord server: https://discord.gg/zSq8rtW
• Please read the Darknet/YOLO FAQ

Using Darknet/YOLO

• Run "darknet help" to get a list of common commands to run.
  • On Linux, if you completed the build and installation steps, all of the Darknet commands should have been installed to /usr/bin/, such as /usr/bin/darknet.
  • On Windows, if you completed the build and installation steps, all of the Darknet commands should have been installed to "C:/Program Files/Darknet/bin/", such as "C:/Program Files/Darknet/bin/darknet.exe".
• Take a look at the Darknet example applications in src-examples/ to see a list of possible commands to run.
• To train your own custom Darknet/YOLO neural network, see the following:
  • the Darknet/YOLO FAQ,
  • the Darknet/YOLO readme in the repo,
  • Stéphane's YouTube tutorials.

Other Darknet/YOLO Projects

• Both the original pjreddie and alexeyab Darknet repos are no longer used. Please update to the new Darknet/YOLO repo.
  • For additional details, see the FAQ: https://www.ccoderun.ca/programming/darknet_faq/#which_repos_to_use_and_avoid.
• DarkHelp is a professional, free, open-source CLI tool and C/C++/Python API to use Darknet/YOLO.
• DarkMark is a free open-source tool for managing Darknet/YOLO project, annotating images, videos, and PDFs, and generating Darknet/YOLO training projects.
• Stéphane's YouTube channel with many Darknet/YOLO "How-To" videos.

Programming With Darknet/YOLO

Darknet/YOLO is not just a CLI tool. When you build Darknet/YOLO, you're building a library you can call from within C applications, C++ applications, or from Python.

A description of the Darknet/YOLO API can be found here: Darknet/YOLO API

There are other language bindings, but those are not maintained within the Darknet/YOLO project. If you own or know of such a project, let us know and we'll link from here to your project.

The Darknet/YOLO Source Code

The Doxygen pages are meant to assist software developers in understanding the internals of the Darknet/YOLO source code.

Things to look up include:

• darknet.h (C API)
• darknet.hpp (C++ API)
• Darknet (C++ namespace)
• Darknet::parse_arguments()
• Darknet::load_neural_network()
• Darknet::free_neural_network()
• Darknet::set_detection_threshold()
• Darknet::set_non_maximal_suppression_threshold()
• Darknet::set_rounded_corner_bounding_boxes()
• Darknet::predict()
• Darknet::annotate()
• ...and many more. See the C++ namespace Darknet for details.

Other useful functions:

• main()
• darknet_fatal_error()
• train_detector()
  • train_network() (CPU only)
  • train_network_waitkey() (1 GPU)
  • train_networks() (multi-GPU)
• Darknet::TimingAndTracking

Configuration files:

• Darknet::CfgFile
  • Darknet::CfgSection
    • Darknet::CfgLine