Skip to content

yunchenlo/Dockerized-YOLO-on-Rpi-Cluster

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

44 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Dockerized-YOLO-on-Rpi-Cluster

Developer: Yun-Chen Lo

Description

Reference Implementation of paper "Distributed Analytics in Fog Computing Platforms Using TensorFlow and Kubernetes"

The below photo is credited to the author of the paper

alt text

Dependencies

Tensorflow 1.1.1

Kubeadm 1.10.2

kubelet 1.10.2

docker 1.13.1

Folder Description

Application/yolo/

single/ //write yolo for one machine

distributed/ //break yolo into three part, which could be handled by different devices

docker-image/ //contains final distibuted yolo version and required Dockerfile

Deployment/

helmChart/yolo/ //contains params to be passed during execution and worker.yaml, master.yaml 1_node.sh 2_node.sh 3_node.sh

Rpi_ENV/

describe my steps to setup a cluster with one pc as master and several Rpis as workers

Experiment/

contains several files that I have experimented before

Results

Because I use a router shared all wifi device in my Lab, therefore the Internet Overhead is big! In the results shown above I just doubled the CPU & Memory Resources that each operator could use on the same Rpi to simulate the 2 devices speed.

References

  1. tensorflow YOLO implementation
  2. How to partition YOLO
  3. How to write Helm diagram

Special Thanks to Hua-Jun Hong's Help to this Project

About

Run distributed, dockerized YOLO on a Raspberry Pi cluster

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published