Skip to content
This repository has been archived by the owner on Aug 12, 2024. It is now read-only.

Latest commit

 

History

History
63 lines (49 loc) · 2.54 KB

README.md

File metadata and controls

63 lines (49 loc) · 2.54 KB

AI on the Edge

Slides

Browse the slides here

Demos

Demo Hardware Software / Docker images
CoreML OSX machine XCode 9.3.1
Embedded Learning Library (Python 3) Raspberry Pi 3 Model B Docker image: lisaong/pi3-miniconda3-ell:2.3.3, default jupyter port: 28888
Keras with TensorFlow backend (Python 3) Raspberry Pi 3 Model B Docker image: lisaong/rpi-keras:1.0, default jupyter port: 38888
Movidius (Python 3) Raspberry Pi 3 Model B, Movidius Neural Compute Stick Docker image: lisaong/rpi-movidius:2.04, default jupyter port: 8888
TensorFlow (Python 2) Raspberry Pi 3 Model B Docker image: lisaong/rpi-docker-tensorflow:tf1.8.0, default jupyter port: 18888
TensorFlow 2.0, Pytorch 1.3 (Python 3.7) Raspberry Pi 3 Model B Docker image: lisaong/rpi-jupyter-tf2-torch-py37:1.0, default jupyter port: 48888, default tensorboard port: 46006

Raspberry Pi 3 Setup

  1. Download and flash a recent Raspbian Lite image onto a 16GB or larger micro SD card (last verified on Stretch). You may also wish to try a DietPi image (last verified on Buster).
  2. Boot up the Raspberry Pi 3
  3. Git clone this repository
  4. Bootstrap git-lfs
cd stackup-workshops/ai-edge/bootstrap
sh ./install_git_lfs_rpi.sh
git lfs pull
  1. Install docker
curl -sSL get.docker.com | sh
sudo usermod -aG docker pi

# IMPORTANT: log out, then log back in again for changes to take effect, then run the next line

sudo systemctl start docker
  1. Launch a demo by navigating to its folder and starting the docker container
cd stackup-workshops/ai-edge/docker/tensorflow
./launch_docker.sh

If all goes well, you should see a Jupyter uri and token. Copy and paste that uri (with token) to your browser on the desktop, and replace localhost with the ip address of the Raspberry Pi 3.

Note that each demo is configured to run in its docker container, which has the necessary environment and tools setup. Mixing and matching containers and demos won't work.

Rendering Presentations

For a better viewing experience, you can render the Jupyter notebook in slide presentation view (reveal.js slideshows):

  1. Install Miniconda
  2. Install RISE
conda install -c damianavila82 rise

From the Jupyter notebook:

  • Alt-r, "Enter/Exit Live Reveal Slideshow
  • Shift-i, Toggle slide
  • Shift-u, Toggle subslide
  • Shift-f, Toggle fragment

Usage