Snapcon-workshop

Object detection and image classification with Snap! and Nvidia Jetson.

Table of content

• Requirements
• Setup
  • Building the Container
  • Running the Container
• Running the program
  • Image Classification
    • Nvidia Jetson
    • Snap!
  • Object detection
    • Nvidia Jetson
    • Snap!

Requirements

• Nvidia Jetson with Jetpack 4.6
• Docker

Setup

Building the Container

Use the docker/build.sh to build docker container.

$ git clone https://github.com/EOLab-HSRW/Snapcon-workshop.git
$ cd Snapcon-workshop
$ docker/build.sh

Running the Container

Use the docker/run.sh to run docker container.

$ docker/run.sh

Running the program

You can run image classification and object detection program.

Image Classification

Run python websocket server in Nvidia Jetson and open classification program in Snap!.

Nvidia Jetson

Run python/Classification.py in Nvidia Jetson.

python3 python/Classification.py

note: If it is the first time running the program, please wait couple of minutes for TensorRT to finish optimizing the network.

Snap!

Open snap/Snap!_with_classification.xml with Snap!.

note: Please use offline version of Snap! with Google Chrome browser to avoid any malfunction.

Write down your Nvidia Jetson ip adress as input for connect block < ws://ip_address:4040 >.

note: You can use ifconfig command in Nvidia Jetson to obtain ip address.

Object detection

Run python websocket server in Nvidia Jetson and open object detection program in Snap!.

Nvidia Jetson

Run python/Detection.py in Nvidia Jetson.

python3 python/Detection.py

note: If it is the first time running the program, please wait couple of minutes for TensorRT to finish optimizing the network.

Snap!

Open snap/Snap!_with_detection.xml with Snap!.

note: Please use offline version of Snap! with Google Chrome browser to avoid any malfunction.

Write down your Nvidia Jetson ip adress as input for connect block < ws://ip_address:4040 >.

note: You can use ifconfig command in Nvidia Jetson to obtain ip address.