Using YOLO models on nvidia jetson

January 21, 2020

YOLO is a highly optimized machine-learning model to recognize objects in videos and images. Running these on your jetson nano is a great test of your board and a bit of fun. To use these awesome models you need to install darknet, the program that runs interference on a video stream from your camera.

What you need:

Jetson Nano Developer Kit with Jetson Nano Developer Kit SD Card image installed
An USB web-camera

Since we have a CUDA gpu on the jetson we should modify the Makefile to enable CUDA and make sure that nvcc is on $PATH. We also want to use OpenCV to make use of our webcam later on.

# cuda/bin is not on PATH by default, add it to .bashrc
echo "export PATH=$PATH:/usr/local/cuda/bin" >> ~/.bashrc
source ~/.bashrc

Now get the source for darknet

git clone https://github.com/AlexeyAB/darknet

Now edit the Makefile, changing GPU=0 to GPU=1, CUDNN=0 to CUDNN=1 and OPENCV=0 to OPENCV=1 and build

cd darknet		
make

Download the YoloV3 and run the model

# Download the pretrained yolo3 model
wget https://pjreddie.com/media/files/yolov3.weights

# run the model 
./darknet detector demo cfg/coco.data cfg/yolov3.cfg yolov3.weights

Voila, here is a demo on vimeo: