fabriziomiano / darknet-docker

Darknet Docker wrapper for all Yolo version

Darknet Docker wrapper

Yolo it's cool, but installing it's a pain. Which OpenCV, which CUDA? This script will make your life a bit easier.

WTF is this?

Docker is the most beautiful thing in the world: you can install anything you want and it will work everywhere (yep, even Darknet).

What is included in this repo

• You have both CPU and GPU verisons.
• Python example for yolov3-tiny (weights included)
• All config and data for darknet are included (if something is missing go here)

Docker installation

CPU only

https://docs.docker.com/install/ Then sudo bash build_cpu.sh

With GPU

https://github.com/NVIDIA/nvidia-docker Then sudo bash build.sh

FAQ

How long does it take to build the docker container?

On my computer (powered by i7 5820k) it took almost an hour.

cv2.imshow doesn't work

Because you have to enable desktop forward, so run xhost + on your host (not the container).

I got CUDA path build errors while compiling the GPU version

Edit build_darknet/Makefile at row 50-51 and specify your CUDA path.

How can I add more Python packages?

Open the Dockerfile and search for

# Install python dependencies
#RUN apt-get install python3-tk -y
#RUN pip3 install matplotlib

Then jst edit anything as you would do in a native machine (not as on Anaconda).