rheartpython/percept-jupyterlab repository overview

⁠JupyterLab running on the Azure Percept DevKit

The purpose of this image is to provide an example of:

• Testing the Eye camera and RTSP stream
• Training a small PyTorch model and testing on a live stream

⁠Requirements

• Azure Percept DevKit - Purchase⁠ (make sure to set up with out-of-box-experience to connect to wifi and assign IP address)
• All default IoT Edge Modules running including the Azure Eye module (for the RTSP stream)
• 3 GB of free space on the device
• [Optional] 512MB swap file (see below)

⁠Prerequisites

The following is required along with "rawTCP" path in URI to access the eye camera RTSP stream/feed.

• Add iptables rule for tcp: sudo iptables -A INPUT -p tcp --dport 8554 -j ACCEPT
• Add to the deployment manifest port bindings for eye module: "8554/tcp":[{"HostPort":"8554"}]

⁠Instructions

SSH into the device with IP address of device. Create the notebooks directory in the base of the user (home/<linux username>). This directory will be mounted inside of the container as well with the docker run command below.

It's a good idea if you have space on the HD to create a 512MB swap file so that JupyterLab has enough memory to run (in case we run out). Refer to Creating a swap file on RedHat⁠.

Start the container in interactive mode (for debugging), mounting a local directory to the container so Jupyter notebooks are saved on the system. Note: if a file exists on the system with the same name as a file in the container, when the container is started, the system file will overwrite the file in the running container. Also, once the container is running, any changes to that file will be saved to the local file system, preserving your work.

sudo docker run -it --rm  -p 8888:8888  -v /home/<linux username>/:/home/tmp --privileged rheartpython/percept-jupyterlab:pytorch-1.8

Once in the container:

mv notebooks tmp/

Now the notebooks and your work will be saved to the local filesystem on the device. Note, any local changes, when rerunning the container, will overwrite what is in the container.

Visit on your dev/local machine that resides on the same LAN or wifi network: http://<ip of your device>:8888

The default password is "microsoft".