Deprecated

Now there are number of commercial products doing this so I've also switched to one of them. :)

Machine Learning Home Security

It's a small project runs surveillance camera image analysis using AWS Rekognition that I wrote for company preso.

Designed to capture email notification from Synology surveillance station and send SMS notification when human movement detected.

It also has MVP frontend web interface to display historic record.

Initial Deployment HOWTO

backend

claudia create --region ap-southeast-2 --handler index.handler --policies policy

Need S3 PUT action trigger as well

ses-mail-parser

claudia create --region us-west-2 --handler index.handler --policies policy

frontend

claudia create --region ap-southeast-2 --handler lambda.handler --deploy-proxy-api --policies policy

Create default collection

aws rekognition create-collection --collection-id "<collection ID>"

Add SES Ruleset with two actions

Recipients
Your email address
Actions
1. S3 Action
Write to S3 bucket <your bucket> under the prefix <prefix>/
2. Lambda Action
Invoke Lambda function <ses-mail-parser> as Event

email bucket should have bucket policy enabling SES write

{
   "Version": "2012-10-17",
   "Statement": [
       {
           "Sid": "GiveSESPermissionToWriteEmail",
           "Effect": "Allow",
           "Principal": {
               "Service": "ses.amazonaws.com"
           },
           "Action": "s3:PutObject",
           "Resource": "arn:aws:s3:::<bucket_name>/*",
           "Condition": {
               "StringEquals": {
                   "aws:Referer": "<AWS ACCOUNT ID>"
               }
           }
       }
   ]
}

cognito integration

Create a user pool and app client. Then OAuth 2.0 set to Implicit grant