INSTALL.md

#########################################################################################################

HOW TO INSTALL YOLO WITH GPU ON LINUX

by Brian Duenas

Tested on Ubuntu 18.04

Graphics card GTX 1060 mobile

I am not responsible for damages to your device use at your own risk. None of this software is mine

I got them from all parts of the web and put them into one guide!

Special thanks to sentex https://pythonprogramming.net/

#########################################################################################################

Follow these steps from "https://pjreddie.com/darknet/yolo/" and test if its working

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

In the darknet directory

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

If make command worked skip these steps

$ sudo apt update $ sudo apt install build-essential #If build essentials doesnt install everything like it did for me in ubuntu $ sudo apt intsall gcc $ sudo apt install g++

Test it out without gpu

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

                          # CONTINUE ONLY IF THE ABOVE WORKS"

Download cuda toolkit as a .run file, cuDNN as a .tgz compressed file

Toolkit: "https://developer.nvidia.com/cuda-downloads" cuDNN: "https://developer.nvidia.com/cudnn"

Give it permissions

$ chmod +x "cuda-toolkit.run-file"

Clean up a couple of things before installing

follow all instructions here: "https://tutorials.technology/tutorials/85-How-to-remove-Nouveau-kernel-driver-Nvidia-install-error.html"

Install all parts of toolkit

$ sudo ./"cuda-toolkit.run-file"

Extract cuDNN files

$ tar zxvf "cudnn-file.tgz"

Go to directory where you extracted cuDNN presumably /Downloads and copy files over

$ cd ~/Downloads $ sudo cp cuda/include/cudnn.h /usr/local/cuda/include $ sudo cp cuda/lib64/* /usr/local/cuda/lib64

Grant permissions

$ sudo chmod a+r /usr/local/cuda/lib64/libcudnn*

Need to export the system path to CUDA elements:

$ sudo nano ~/.bashrc

Go to the very end of this file, and add:

$ export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib64 $ export CUDA_HOME=/usr/local/cuda $ export PATH=$PATH:/usr/local/cuda/bin

Reload paths

$ source ~/.bashrc

For python usage later on add these

$ sudo nano /etc/enviornment

Reload changes

$ sudo ldconfig

Now edit the 'Makefile' in darknet

$ sudo nano Makefile -> GPU=1 -> CUDNN=1

Restart computer

$ sudo reboot

Head back to darknet directory

$ make clean $ make

If the above did not work do the same steps for '/.profile' as you did for '/.bashrc'