sprcoder/ImageRecognition

Java Maven Project running in multiple EC2 instances in AWS that uses various services such as SQS, Rekognition, S3 Bucket and EBS to perform image recognition.

AWS_Simple_ImageRecognition

EC2 Instance Setup

1. Created 2 EC2 instances in AWS, EC2-A & EC2-B
2. Each Instace would have the following configuration while creating.

• OS -> Amazon Linux 2 AMI 64 bit(x86)

• Intance Type -> Selected the t2.micro that has free tier eligibility

• Key Pair -> Created a new pair for using with my computer as MyMacConnect

• Network Settings -> Created a setting with default values for ssh, http, https

• Used the same setting for the 2 EC2 instances

• Configure Storage -> Used the default free tier storage that was available for both the instances

• Launch Instance -> With the below configurations, launched the EC2 instance

Running the Application

1. Requires SQS setup before starting the application

• Create a FIFO Queue for the instances to communicate

2. EC2-A Instance

• Connect to the EC2-A Instance using SSH

• In the instance run aws configure

  • This would ask for the access key, id and session details that are available in the Learners Lab, AWS Details.

• Once AWS is configured install java, I have used jdk-11.0.16.1_linux-x64_bin.rpm

• After java installation, install maven, I have used apache-maven-3.9.0. Provided maven bin path to $PATH variable in linux to finalize installation.

• Create a maven project with the following code

   mvn archetype:generate \
  -DarchetypeGroupId=software.amazon.awssdk \
  -DarchetypeArtifactId=archetype-app-quickstart \
  -DarchetypeVersion=2.18.16

• Use the following configuration while creating the maven project

- Below is the structure of the generated project

├── src
│   ├── main
│   │   ├── java
│   │   │   └── package
│   │   │       ├── App.java
│   │   │       ├── DependencyFactory.java
│   │   │       └── Handler.java
│   │   └── resources
│   │       └── simplelogger.properties
│   └── test
│       └── java
│           └── package
│               └── HandlerTest.java

• App.java: main entry of the application
• DependencyFactory.java: creates the SDK client
• Handler.java: you can invoke the api calls using the SDK client here.

3. Handler.java holds the main code for my application, which are the following

• S3 bucket initialization and accessing the S3Object. Used the bucket provided in the assignment njit-cs-643.
• Passed the image value from S3Object to AWS Rekognition method that would identify the label of the image.
• Image index that is identified as car with confidence more than 90 is sent as SQS message using the sqsClient & sendMessage to the queueurl of A-B.fifo.

4. ImageRecog.java was created to hold the AWS Rekognition method.

• This method creates a rekognitionClient and use the DetectLabelsRequest object to identify the labels in the image.

5. Run the program by cleaning the mvn package and executing the package:

mvn clean package
mvn exec:java -Dexec.mainClass="org.example.App"

• Please refer to output_a.txt in ec2_a_imgrecog.zip/imgRecog/ for the results of execution.

6. EC2-B Instance

• For the initial setup of the instance follow the step 2
• Handler.java holds the main code for EC2-B instance application, which are the following
  • Checks for SQS messages from the FIFO queue A-B.fifo using ReceiveMessageRequest & sqsClient. Tries to get all messages with 20 attemps(As a exit condition, if no message is received from SQS)
  • Each request from a SQS message waits for 5 seconds to get the message.
  • Once message is received, index from the body of the message is retrieved. And the message is deleted using DeleteMessageRequest.
  • Passing the index of the image to AWS Rekognition method with the bucketname would retieve a list of texts that is found in the image.
  • The image information along with the text identified from the image is written into a file.
• TextRecog.java was created to hold the AWS Rekognition method.
  • This method would get the image name, bucket name and using the rekognitionClient & DetectTextRequest identifies the text present in the image.
• Follow the step 5 to run the application
  • Please refer to output_b.txt in ec2_b_imagerecog_textrecog.zip/imageRecog/ for the results of execution.