JetRacer

Requirements

• Jetson Nano 2GB Developer Kit
  • GPU : 128-core NVIDIA Maxwell architecture GPU
  • CPU : Quad-core ARM® Cortex®-A57 MPCore processor
• Jetracer
  • Cam : Sony IMX219 Sensor, Resolution : 3280 x 2464
• micro SD card (64GB~)
• balenaEtcher

Installation

1. Format micro SD card before using.
2. Jetson Nano Image

• Download jetson-nano-2gb-sd-card-image & Uncompress
• Write Jetson Nano Image in micro SD card using Etcher program.
• Put SD card into Jetson nano's SD card slot, and connect to the monitor using HDMI port. Finish system configuration

3. JetRacer Image

• Download JetRacer image & Uncompress
• Write JetRacer Image in micro SD card using Etcher program.
• Put SD card into Jetson nano's SD card slot, and connect to the monitor using HDMI port. Finish system configuration

(+)

• To activate GUI, use sudo systemctl set-default graphical.target command

Connect to JetRacer

1. Connect JetRacer to WIFI network
2. Check ip address of Wlan0 interface using Jetson Nano's piOLED or ifconfig command
3. Connect to JetRacer wirelessly http://<ip address>:8888
4. Initial password is jetbot. Run Jupyter Lab!

• For JetRacer package update, execute following command.

cd jetracer
git checkout master
sudo python3 setup.py install
sudo reboot

Load trained model

Upload mean image

/datasets/mean_image.npy

Upload trained model

/checkpoints/resnet18.pth

Run

Code : JetRacerRun.ipynb

1. Import nvidia_racecar, csi_camera by Cell 1
2. Create & load model by Cell 4
3. Read image from the camera, preprocess the image
4. Compute current pose using your model

Demo

Ewha W. University, Asan Building, 2F