search

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.8k stars 7.97k forks source link

Configuring incomplete, errors occurred! #4385

Open savinoda opened 5 years ago

savinoda commented 5 years ago

I'm trying to compile darknet on Windows 10. I installed all the requirements: -CMAKE 3.16.0-rc4 -CUDA 10.2.89 -opencv-4.1.2-vc14_vc15 -cudnn-10.1 -GPU 1050ti -MSVC 2015/2017/2019 (community 893465157.1564759710)

When compiling the source files with CMake gui I have the following error: Configuring incomplete, errors occurred!

And in the error log file: See also "C:/Users/AAA/Desktop/darknet/CMakeFiles/CMakeOutput.log". The only text is The system is: Windows - 10.0.17763 - AMD64

Am I screwing up anything with dependences or what?

AlexeyAB commented 5 years ago

@cenit Hi, do you know what could be the reason?

cenit commented 5 years ago

@savinoda can you please share the file CMakeOutput.log?

Also, please try building Darknet from a powershell, using the script .\build.ps1 It is much more reliable than cmake gui

savinoda commented 5 years ago

As I said the file has got this content: The system is: Windows - 10.0.17763 - AMD64 In the documentation it is said that the recommended approach is the one with cmake gui... I will try to build with the script, but maybe there is an error in the file and it would be nice to fix the problem

savinoda commented 5 years ago

Compilation was successful with build.ps1 but I could not compile with cuda support for openCV (I left the whole night this command running .\vcpkg install pthreads opencv[cuda,ffmpeg] with no result and I had to fallback to .\vcpkg install pthreads opencv[ffmpeg]). If I can do anything useful to solve the problem with CMAKE Gui please tell me. Thanks for your help

cenit commented 5 years ago

building opencv with cuda support is very time consuming. On a very good machine it takes up to 3 hours, it wouldn't be strange for it to take more than 12 hours. Unfortunately I need more information to help you with CMake Gui. Any screenshot or log to diagnose better the problem would be nice. Is there any CMakeError.log in the folder? Anything interesing anywhere?

savinoda commented 5 years ago

@cenit Thanks for the answer. I Attached two pictures of the gui error and the log file. If you need other please ask.

error error1

Installing open cv without cuda is also the reason why I have the following error? SED SDK not enabled since it requires CUDA. buildps1

Then when training with the following command, it only uses CPU and shows me nan AVG loss. I used the script in the description to mark my images (22 jpg from 2 classes) and modified all the config files accordingly command

AlexeyAB commented 5 years ago

@savinoda In general it isn't necessary to compile OpenCV with CUDA, so you can just do .\vcpkg install pthreads opencv[ffmpeg]

Darknet will still use CUDA GPU.