Skip to content

The Pico Project

Object Detection & Text Analytics using Deep Learning Made Simple using Docker, Apache Kafka, IoT & Amazon Rekognition System

image

Table of Contents

  1. Intent
  2. Hardware
  3. Software
  4. Activating Raspberry Pi Camera
  5. Setting up Docker on Raspberry Pi
  6. Preparing Your Raspberry Pi
  7. Using Raspberry Pi Imager
  8. Installing Docker 19.03 on Raspberry Pi
  9. Setting up Apache Kafka running inside Docker container on AWS Cloud EC2 Instance
  10. Building up First Node using Docker Machine
  11. Building Apache Kafka on 2-Node Docker Swarm Cluster
  12. Setting up Pico
  13. Running Consumer Scripts on AWS Cloud Instance
  14. Running Producer Script on Raspberry Pi
  15. References & Resources

Intent:

The intention with this project is to showcase how easy it is to implement object detection and analytics using Docker containers.Imagine you are able to capture live video streams, identify objects using deep learning, and then trigger actions or notifications based on the identified objects - all using Docker containers. With Pico, you will be able to setup and run a live video capture, analysis, and alerting solution prototype.

image

A camera surveils a particular area, streaming video over the network to a video capture client. The client samples video frames and sends them over to AWS, where they are analyzed and stored along with metadata. If certain objects are detected in the analyzed video frames, SMS alerts are sent out. Once a person receives an SMS alert, they will likely want to know what caused it. For that, sampled video frames can be monitored with low latency using a web-based user interface.

The Pico framework uses Kafka cluster to acquire data in real-time. Kafka is a message-based distributed publish-subscribe system, which has the advantages of high throughput and perfect fault-tolerant mechanism. The type of data source is the video that generated by the cameras attached to Raspberry Pi.

image

Hardware

Items Link Reference
Raspberry Pi 3 Model B Buy Buy
Raspberry Pi Infrared IR Night Vision Surveillance Camera Module 500W Webcam Buy Buy
5MP Raspberry Pi 3 Camera Module W/ HBV FFC Cable Buy Buy

image

Software

  1. Raspberry Pi OS
  2. Docker 19.03.x
  3. Python
  4. Amazon Cloud Subscription
  5. AWS Rekognition Service

Activating Raspberry Pi Camera Module

To configure the camera Interface, run the below command as sudo or root user:

$ sudo raspi-config

It will open up command-line UI window, choose Interfacing , select Camera and enable it. Save and exit the CLI window.

You will also need to load the required driver “bcm2835-v412” to make your camera module work. If you miss this step, you will end up seeing a blank screen even though the application comes up without any issue.

# sudo modprobe bcm2835-v4l2

image

Setting up Docker on Raspberry Pi

1. Preparing Your Raspberry Pi

Raspberry Pi OS (previously called Raspbian) is an official operating system for all models of the Raspberry Pi. We will be using Raspberry Pi Imager for an easy way to install Raspberry Pi OS on top of Raspberry Pi:

Visit this link and download Raspberry Pi OS by running the below CLI:

image

In case you are in hurry, just run the below command and you should be good to go:

wget https://downloads.raspberrypi.org/raspios_full_armhf_latest

2. Using Raspberry Pi Imager

Next, we will be installing Raspberry Pi Imager. You can download via https://www.raspberrypi.org/blog/raspberry-pi-imager-imaging-utility/

image

All you need to do is choose the right operating system and SD card, and it should be able to flash OS on your SD card.

image

Click “Write” and it’s time to grab a coffee.

image

Once the write is successful, you can remove the SD card from card reader and then insert it into Raspberry Pi SD card slot.

image

$ssh pi @192.168.1.4
pi@raspberrypi:~ $ uname -arn
Linux raspberrypi 4.19.118-v7+ #1311 SMP Mon Apr 27 14:21:24 BST 2020 armv7l GNU/Linuxpi@raspberrypi:~ $

3. Installing Docker 19.03 on Raspberry Pi

sudo curl -sSL https://get.docker.com/ | sh

Verifying Docker Binaries

pi@raspi2:~ $ docker version
Client: Docker Engine - Community 
Version:           19.03.4 
API version:       1.40 
Go version:        go1.12.10 
Git commit:        9013bf5 
Built:             Fri Oct 18 16:03:00 2019 
OS/Arch:           linux/arm 
Experimental:      false

Server: Docker Engine - Community Engine:  
Version:          19.03.8  
API version:      1.40 (minimum version 1.12)  
Go version:       go1.12.17  
Git commit:       afacb8b  
Built:            Wed Mar 11 01:29:22 2020  OS/Arch:          linux/arm  Experimental:     false 
containerd:  Version:          1.2.10  
GitCommit:        b34a5c8af56e510852c35414db4c1f4fa6172339 
runc:  Version:          1.0.0-rc8+dev  
GitCommit:        3e425f80a8c931f88e6d94a8c831b9d5aa481657 docker-init:  
Version:          0.18.0  
GitCommit:        fec368
pi@raspi2:~ $

Setting up Apache Kafka running inside Docker container on AWS Cloud EC2 Instance

Pre-requisites:

  • Docker Desktop for Mac or Windows
  • AWS Account ( You will require t2.medium instances for this)
  • AWS CLI installed
  • Docker Machine installed

Adding Your Credentials:

[Captains-Bay]🚩 >  cat ~/.aws/credentials
[default]
aws_access_key_id = XXXA 
aws_secret_access_key = XX

Verifying AWS Version

[Captains-Bay]🚩 >  aws --version
aws-cli/1.11.107 Python/2.7.10 Darwin/17.7.0 botocore/1.5.70
Setting up Environmental Variable

[Captains-Bay]🚩 >  export VPC=vpc-ae59f0d6
[Captains-Bay]🚩 >  export REGION=us-west-2a
[Captains-Bay]🚩 >  export SUBNET=subnet-827651c9
[Captains-Bay]🚩 >  export ZONE=a
[Captains-Bay]🚩 >  export REGION=us-west-2

Building up First Node using Docker Machine

[Captains-Bay]🚩 >  docker-machine create  --driver amazonec2  --amazonec2-access-key=${ACCESS_KEY_ID}  --amazonec2-secret-key=${SECRET_ACCESS_KEY} --amazonec2-region=us-west-2 --amazonec2-vpc-id=vpc-ae59f0d6 --amazonec2-ami=ami-78a22900 --amazonec2-open-port 2377 --amazonec2-open-port 7946 --amazonec2-open-port 4789 --amazonec2-open-port 7946/udp --amazonec2-open-port 4789/udp --amazonec2-open-port 8080 --amazonec2-open-port 443 --amazonec2-open-port 80 --amazonec2-subnet-id=subnet-72dbdb1a --amazonec2-instance-type=t2.micro kafka-swarm-node1

Listing out the Nodes

[Captains-Bay]🚩 >  docker-machine ls
NAME                ACTIVE   DRIVER      STATE     URL                         SWARM   DOCKER     ERRORS
kafka-swarm-node1   -        amazonec2   Running   tcp://35.161.106.158:2376           v18.09.6   
kafka-swarm-node2   -        amazonec2   Running   tcp://54.201.99.75:2376             v18.09.6

Initialiating Docker Swarm Manager Node

ubuntu@kafka-swarm-node1:~$ sudo docker swarm init --advertise-addr 172.31.53.71 --listen-addr 172.31.53.71:2377
Swarm initialized: current node (yui9wqfu7b12hwt4ig4ribpyq) is now a manager.

To add a worker to this swarm, run the following command:

    docker swarm join --token SWMTKN-1-xxxxxmr075to2v3k-decb975h5g5da7xxxx 172.31.53.71:2377

To add a manager to this swarm, run 'docker swarm join-token manager' and follow the instructions.

Adding Worker Node

ubuntu@kafka-swarm-node2:~$ sudo docker swarm join --token SWMTKN-1-2xjkynhin0n2zl7xxxk-decb975h5g5daxxxxxxxxn 172.31.53.71:2377
This node joined a swarm as a worker.

Verifying 2-Node Docker Swarm Mode Cluster

ubuntu@kafka-swarm-node1:~$ sudo docker node ls
ID                            HOSTNAME            STATUS              AVAILABILITY        MANAGER STATUS      ENGINE VERSION
yui9wqfu7b12hwt4ig4ribpyq *   kafka-swarm-node1   Ready               Active              Leader              18.09.6
vb235xtkejim1hjdnji5luuxh     kafka-swarm-node2   Ready               Active                                  18.09.6

Installing Docker Compose

curl -L https://github.com/docker/compose/releases/download/1.25.0-rc1/docker-compose-`uname -s`-`uname -m` -o /usr/local/bin/docker-compose
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100   617    0   617    0     0   2212      0 --:--:-- --:--:-- --:--:--  2211
100 15.5M  100 15.5M    0     0  8693k      0  0:00:01  0:00:01 --:--:-- 20.1M

root@kafka-swarm-node1:/home/ubuntu/dockerlabs/solution/kafka-swarm# chmod +x /usr/local/bin/docker-compose

ubuntu@kafka-swarm-node1:~/dockerlabs/solution/kafka-swarm$ sudo docker-compose version
docker-compose version 1.25.0-rc1, build 8552e8e2
docker-py version: 4.0.1
CPython version: 3.7.3
OpenSSL version: OpenSSL 1.1.0j  20 Nov 2018

Building Apache Kafka on 2-Node Docker Swarm Cluster

Apache Kafka is an open-source stream-processing software platform developed by LinkedIn and donated to the Apache Software Foundation. It is written in Scala and Java. The project aims to provide a unified, high-throughput, low-latency platform for handling real-time data feeds.

Apache Kafka is a distributed, partitioned, and replicated publish-subscribe messaging system that is used to send high volumes of data, in the form of messages, from one point to another. It replicates these messages across a cluster of servers in order to prevent data loss and allows both online and offline message consumption. This in turn shows the fault-tolerant behaviour of Kafka in the presence of machine failures that also supports low latency message delivery. In a broader sense, Kafka is considered as a unified platform which guarantees zero data loss and handles real-time data feeds.

Cloning the Repository

git clone https://github.com/ajeetraina/developer/solution/iot/ai/pico
cd pico/kafka/

Using docker stack deploy to setup 3 Node Kafka Cluster

docker stack deploy -c docker-compose.yml mykafka

By now, you should be able to access kafka manager at https://:9000

Adding a cluster

  • Cluster Name = pico (or whatever you want)
  • Cluster Zookeeper Hosts = zk-1:2181,zk-2:2181,zk-3:2181
  • Kafka Version = leave it at 0.9.01 even though we're running 1.0.0
  • Enable JMX Polling = enabled

Adding a Topic

Click on Topic on the top center of the Kafka Manager to create a new topic with the below details -

  • Topic = testpico
  • Partitions = 6
  • Replication factor = 2

which gives an even spread of the topic across the three kafka nodes.

While saving the settings, it might ask to set minimal parameter required. Feel free to follow the instruction provided.

Setting up Pico

Running Consumer Scripts on AWS Cloud Instance

Run the below Docker container for preparing environment for Consumer scripts

docker run -d -p 5000:5000 ajeetraina/opencv4-python3 bash

Open up bash shell inside Docker Container

docker exec -it <container-id> bash

Remove the existing Pico directory

rm -fr pico

Cloning the fresh Repository

#git clone https://github.com/ajeetraina/developer/
cd solutions/iot/ai/pico

Locating the right consumer scripts

You will need 2 scripts - Image Processor and Consumer

cd pico/deployment/objects/

Execute Image processor Script

This script is placed under hhttps://github.com/ajeetraina/developer/edit/master/solutions/iot/ai/pico/blob/master/deployment/objects/image_processor.py location. Before you run this script, ensure that it has right AWS Access Key and Broker IP address

python3 image_processor.py

Open up new bash again

docker exec -it <container-id> bash

Exexute Consumer Script

This script is placed under https://github.com/ajeetraina/developer/edit/master/solutions/iot/ai/pico/blob/master/deployment/objects/ directory. Before you run this script, ensure that it has right Broker IP address

python3 consumer.py

Running Producer Script on Raspberry Pi

Clone the Repository

git clone https://github.com/ajeetraina/developer
cd developer/solutions/iot/ai/pico

Locating Producer Script

cd pico/deployment/objects/

Edit producer_camera.py script and add the proper IP address for the kafka broker:

brokers = ["35.221.213.182:9092"]

Installing Dependencies

apt install -y python-pip libatlas-base-dev libjasper-dev libqtgui4 python3-pyqt5 python3-pyqt5 libqt4-test
pip3 install kafka-python opencv-python pytz
pip install virtualenv virtualenvwrapper numpy

Execute the script

python3 producer_camera.py

Please Note: This script should be run post the consumer scripts (Image_Processor & Consumer.py) is executed

Testing Object Detection

Sequence of Scripts Execution

Pre-requisite:

  • Ensure that you have followed all the above steps.
  • Ensure that Docker Swarm is up and running on AWS Cloud

Sequence:

  • First run the Image_Processor Script on AWS Instance
  • Then run the Consumer.py Script on AWS Instance
  • Finally, run the Producer_camera.py script on Pi

Place an object in front of camera module and watch out for both text as well as object detection under http://broker-ip:5000

image

References & Resources