Kickstart-Jetson

Kickstart Jetson STEM Course Bootstraping the Jetson nano with Darknet Yolo

Step 0. Download the image for NVIDIA Jetson (already prepared for you)

Jetson Nano Developer Kit SD Card Image https://developer.nvidia.com/embedded/downloads and burn ISO using balenaEtcher

Step 1. Insert SD Card, connect Camera, fan, etc

boot up the Jetson nano Ensure the JP48 is connected, connect the 5V 4A power

Step 2. Clone my opensource repo

This automatically do some magic stuffs

git clone https://github.com/carryai/kickstart-jetson.git
cd kickstart-jetson

Step 3. Configure network, make a hotspot for easy control

Create Wi-Fi hotspot you might need to change the code to change ssid and password

./configure-network.sh

	ascii
SSID	i_am_jetson
Password	jetsonnano

Step 4. Install required software and libraries

./requirements.sh

Step 5. Back to your computer, prepare Visual Studio Code

c:\Users**(your_username)**.ssh\config

Host Jetsonnano-192.168.2.1
	HostName 192.168.2.1
	User jetsonnano
	ForwardAgent yes
	StrictHostKeyChecking no

Step 6. Download Yolo

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

Step 7. Edit Makefile

nano Makefile

Change the following contents of the Makefile

GPU=1
CUDNN=1
OPENCV=1
LIBSO=1

Change the following contents of the Makefile, uncomment the lines

#uncomment the line
ARCH= -gencode arch=compute_53,code=[sm_53,compute_53]

Change the the environment file, ~/.bashrc

nano ~/.bashrc

add the following into the end of ~/.bashrc

export CUDA_VER=10.0
export PATH=${PATH}:/usr/local/cuda/bin
export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/usr/local/cuda/lib64

activate ~/.bashrc

source ~/.bashrc

Step 8. Compile the program

make -j4

Step 9. Download models & weights

cd ~/darknet wget https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-tiny.weights -O yolov4-tiny.weights

wget https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-csp.weights -O yolov4-csp.weights

Step 10. Your first inference!

./darknet detector test ~/darknet/cfg/coco.data ~/darknet/cfg/yolov4-csp.cfg ~/darknet/yolov4-csp.weights data/person.jpg

Step 11. Lets test with USB Camera

cd ~/darknet
export DISPLAY=:1

./darknet detector demo  ~/darknet/cfg/coco.data  ~/darknet/cfg/yolov4-csp.cfg   ~/darknet/yolov4-csp.weights    -thresh 0.50 -ext_output \
'v4l2src io-mode=2 device=/dev/video0 ! image/jpeg, width=1920, height=1080, framerate=30/1 ! jpegdec ! video/x-raw !  nvvidconv ! video/x-raw(memory:NVMM), format=(string)I420 ! nvvidconv ! video/x-raw, format=(string)BGRx ! videoconvert ! video/x-raw, format=BGR ! appsink max-buffers=1 drop=true sync=false'  

Step 12. Take Pictures and build your own training

Let's take some time and allow me to explain what are the important things you need to consider when doing your training

Step 13. Train your weight

However, training is almost impossible to be done on jetson nano, lets do it on cloud, we will use Google Colab

 ./darknet detector train "/training-data/face_mask/obj-google.data"  "/training-data/face_mask/yolov3-tiny-prn-832.cfg"  "/training-data/face_mask/yolov3-tiny-prn-832_last.weights" -dont_show

---

---

Other useful scripts

GPIO

See https://www.jetsonhacks.com/2019/06/07/jetson-nano-gpio/ See https://github.com/NVIDIA/jetson-gpio

##　Running Jetson-IO After the setup, we’re ready to go:

sudo /opt/nvidia/jetson-io/jetson-io.py

Manage Autostart

Using Linux rc.local https://www.linuxbabe.com/linux-server/how-to-enable-etcrc-local-with-systemd https://vpsfix.com/community/server-administration/no-etc-rc-local-file-on-ubuntu-18-04-heres-what-to-do/

sudo vim /etc/rc.local
#allow sending serial
if [ -f "/dev/ttyTHS1" ]; then
        chmod 666 /dev/ttyTHS1
fi

if [ -f "/dev/ttyUSB0" ]; then
        chmod 666 /dev/ttyUSB0
fi

Jetson Nano 3D printer samples

https://www.thingiverse.com/thing:3603594 https://www.thingiverse.com/thing:4082134 https://www.prusaprinters.org/prints/4689-jetson-nano-case https://www.prusaprinters.org/prints/1420-nvidia-jetson-nano-case-nanomesh-mini

Finally

./start.sh

Save as a copy in Drive

https://colab.research.google.com/drive/1czTxmIIcMkqRdgbsTOOmO4ecBeFIrKP8

Connections

How to connect

If you find that the image captured is red. You can try to download .isp file and installed:

https://www.waveshare.com/wiki/IMX219-160_Camera

wget https://www.waveshare.com/w/upload/e/eb/Camera_overrides.tar.gz
tar zxvf Camera_overrides.tar.gz 
sudo cp camera_overrides.isp /var/nvidia/nvcam/settings/
sudo chmod 664 /var/nvidia/nvcam/settings/camera_overrides.isp
sudo chown root:root /var/nvidia/nvcam/settings/camera_overrides.isp
rm Camera_overrides.tar.gz 
rm camera_overrides.isp