Before following along on this tutorial refer to this page to check if you have set up your Jetson Nano properly. If not this will lead to a lot of errors in the future steps of this turorial:
https://github.com/Sanjiv-B-N/jetson-nano-setup
we need darkent to run tinyYOLOv4 on the Jetson Nano. Run the following on the command line:
git clone https://github.com/AlexeyAB/darknet.git
cd darknet
sudo nano Makefile
Now the makefile opens. Make the following changes to the make:
GPU=1
CUDNN=1
OPENCV=1
After that save the Makefile and run the following command to install darknet:
make
We have trained the tinyYOLOv4 on the Elevator button dataset. I have uploaded the config file and the retrained weights in this github. Download them and place them in your darknet directory. Now run this command to have object detection on the Jetson Nano! run this on the commandline in th darkenet directory:
./darknet detect yolov4-tiny-custom.cfg yolov4-tiny-custom_best.weights image.jpg
Here on this line add the correct paths to the weigths the config file and test image and then you'll get the output!