⚒️ Under development ⚒️

A fully dockerized, heterogenous edge computing framework, supporting a range of CPU architectures and machine learning applications.

Table of Contents

1. The Applications
2. Docker image
3. Setup
4. Usage
5. Project Tree
6. Miscellaneous

The Applications

• DeepSpeech, an embedded speech-to-text engine using Tensorflow
• FaceNet, a face recognizer using Tensorflow
• LaneNet, a deep neural network for real time lane detection using Tensorflow
• RETAIN, an interpretable predictive model for healthcare applications using Theano

More applications can be added relatively easily by installing the prerequisites in the docker image and properly configuring the application itself and the apps/start.sh file.

Docker Image

A docker image is utilized for easy & fast deployment on different platforms: Docker Repository

Pull it with docker pull uphilld/edgebench:<platform> or let the launcher do it for you!

Currently supported platforms: arm32v6, arm32v7, arm64v8, amd64

Setup

Run the following command:

git clone https://github.com/UphillD/edgebench && \
cd edgebench && \
git submodule init && \
git submodule update && \
./launcher.sh prepare

or as a single line:

git clone https://github.com/UphillD/edgebench && cd edgebench && git submodule init && git submodule update && ./launcher.sh prepare

Usage

Start the launcher:

Interactive menu: ./launcher.sh

Interactive shell: ./launcher.sh explore

Direct call: ./launcher.sh <command> <arguments>

Project Tree

$root
├ algo 🔒
├ apps
│    ├ deepspeech
│    ├ facenet
│    ├ lanenet
│    ├ retain
│    ├── settings.cfg
│    └── start.sh
├ build
├ data
│    ├ payloads
│    └ models
├ docs
├ scripts
└ workers

The algo folder is omitted for obvious reasons.

The data folder is also omitted, models and payloads can be downloaded through the appropriate menu selection.

Miscellaneous

Shield badges provided by Shields.io.

⇯ Back to Top