Darknet YOLOv4 Object Detection Tutorial for Windows 10 on Images, Videos, and Webcams

YOLOv4 tutorial to build Darknet YOLOv4 object detection model on Windows 10 to achieve real-time object detection on images, videos, and webcam. In this YOLOv4 tutorial, you will learn to compile Darknet YOLOv4 on your local machine with OpenCV and GPU acceleration.

#TheCodingBug
#YOLOv4
#Darknet
---------------------------------------------
► Time Stamps:
Introduction: (0:00)
Prerequisite: (0:21)
Download Darknet: (03:31)
Copy cuDNN and OpenCV Files: (3:55)
Build Darknet using Visual Studio: (4:50)
Object Detection on Images: (8:53)
Object Detection on Videos: (9:48)
Object Detection on Webcams: (10:34)
---------------------------------------------
► Links:
---------------------------------------------
► Commands:

---------------------------------------------
► My Other Tutorials:
---------------------------------------------
---------------------------------------------
DISCLAIMER: Links included in this description might be affiliate links. If you purchase a product or service with the links that I provide I may receive a small commission. There is no additional charge to you!

[louisfain]

gosh i wish i could give you a hug right now. you're the champion! you covered what other youtubers (i won't name names) have missed and saved me from hours of aimlessly fixing things.

[uncleanggi]

Thank you for your great tutorial, now I can sleep after learning this. Haha.

However, allow me to add some solution, if your Cuda is 11.1 or above. In the first build, I encountered error "nvcc fatal : Unsupported GPU architecture 'compute_30'", after exploring on stackoverlow, it seems Cuda 11.0 above is not support compute_30, so, it should be change to compute_80. You can change it by opening make file on darknet folder, and change this line

From this:
ARCH= -gencode arch=compute_50, code=[sm_50, compute_50] \
-gencode arch=compute_52, code=[sm_52, compute_52] \
-gencode arch=compute_61, code=[sm_61, compute_61]

to this:

ARCH= -gencode arch=compute_80, code=sm_80 \
-gencode arch=compute_52, code=sm_52 \
-gencode arch=compute_61, code=sm_61

And then, before you build yolo_cpp_dll.vcxproj, make sure you already paste getopt.c dan getopt.h from to

and the last thing, before you build darknet.sln, right click darknet, go to Project Properties → Configuration Properties → C/C++ → General → Additional Include Directories.
Add the path to the folder containing 'getopt.h' from

Also don't remove compute 80 in your property, when you build the darknet in Vc 2019.

It will resolved the issue, if you build it with Higher GPU and Cuda 11.1 above

Remember, Always using VC 2019!

[dompower]

Excellent tutorials.. It's fast however very much understandable.... I like to see more projects from you..

[jagdishgoklani]

I am getting this error while building yolo_cpp_dll. MSB4019 The imported project "C:\Program Files (x86)\Microsoft Visual 10.1.props" was not found. Confirm that the expression in the Import declaration "C:\Program Files (x86)\Microsoft Visual 10.1.props" is correct, and that the file exists on disk.

I created darknet folder in D Directory. What to do

[mallasudhakara]

I am getting comute30 and compute70 errors sir, please help how to fix it

[vhalwkd]

Thank you so much for your awesome explanation!

[dwikirahman]

i have cuda version 11.2, and when i build in visual studio, it says error about the SDK, do you know why?

[canozkan]

Last version of cudnn doesn't have cudnn64_7.dll, what should we do about that ?

[suminh]

This video was very helpful, it worked for me, thanks a lot. Vietnamese with love

[nugratasik]

Hello thanks for great tutorial! I've made it to the end, but how can I build my custom train with the darknet?
Look forward hearing from you soon!

[deepmodi]

Nice Tutorial must watch it.
Thank You.

[borischang]

In last step I want build into darknet.exe in Visual studio, it would fail, and than it must save as Unicode, but what should I do?

[jaso]

help...problem building darknet in the same versions used in the error's....e.g :Error calling a __host__ function("__floorf") from a __device__ is not allowed manny were in the following files : 'activation_kernals.cu' , 'blas_kernals.cu' and 'CUDA 4 warnings:
"variable "MISH_THRESHOLD" was declared but never referenced" file: activation_kernals.cu
"variable "out_index" was declared but never referenced" file:blas_kernals.cu
"variable "step" was set but never used" file:blas_kernals.cu,
"variable "stage_id" was declared but never referenced" file:blas_kernals.cu.

[riteshananth]

I get Video-stream stopped error when i run it on videos, what could be the issue?
I can run it on images and web-cam, but for videos i am unable to

[hildanfebianto]

in the previous video "opencv dnn module" i have successfully saved video output. how to saved video output in this tutorial?

[ronaktawde]

I have following configuration i.e.
windows 10
GPU - NVIDIA 1660ti
Cuda 10.1 cudnn 7.5.1
can I install opencv 4.5.2 for yolo v3 and yolo v4?

[arunsoftwareeng]

In build folder, opencv_world440.dll file is missing in my build folder. how can i generate it?

[sefadogan]

Hello, with which command can I detect object on all photos in a particular folder in yolov4? I want it to be done one by one in all the photos one after the other.

[williamwilliam]

Hello, i succeeded on running yolo with webcam, but my stream resolution is stuck at 640 x 360, do you know how to change the resolution to 1920 x 1080? thank you for your help, the tutorial was really helpful to me

[daryrafi]

I don't have file cudnn64_7 but cudnn64_8 is it okay?