Detect and Locate Stranger Intrusion


The motion notification from normal security cameras such as Samsung is annoying. I don't want a notification when my pet or kids move around the house. I only want an alarm notification when a stranger comes to my house, and I want to see the face of the stranger immediately without a playback of the video recording.

In some scenarios, stranger intrusion notification in real-time is not good enough. For example, security guards need the accurate location of the stranger to take immediate actions. Therefore the notification should have stranger's location information.

What it does

In this project, we send realtime SMS/email notification of stranger's location and photo using Deeplens and Cisco CMX WiFi location service.

Created By: Gracia Wang and Lucy Huang

How we built it

Below is the work flow we used to build the system:

  1. Build lambda function to detect face object, send face image and Deeplens MAC address to S3
  2. Deploy above lambda function and face detection model to Deeplens
  3. Build lambda function to call Rekognition API to compare detected face with faces in collection named "family", check if it's a known face in the collection or not. If an unknown stranger face detected, publish message to SNS topic, which trigger an email notification. Setup this lambda function trigger by S3 new image upload event
  4. Setup SNS topic for email message notification
  5. To obtain location information of stranger, we imported testing map and location hierarchy into CMX account, placed APs in the map in CMX cloud service ( In the meanwhile, we enhance lambda function in 3, to read MAC address from S3 file, and call CMX client API with MAC address as query parameter to obtain Deeplens device location


The DeepLens device OS is not stable. The software automatically upgraded and it introduced regression on the device. Synced with AWS DeepLens support team. Eddie and the team helped solve the issue with support overnight. Finally I'm able to test out the project on the device. Appreciate the professional online support from AWS team!

Accomplishments that we're proud of

It's unbelievable we were able to accomplish a deep learning project within a couple of weeks! With limited knowledge of deep learning, we were able to achieve using it to solve the real life problem in such a short time. Moreover, we were inspired by the DeepLens, and we came up with many interesting ideas when we brainstormed in the beginning.

Even my 11 year old daughter showed interest in DeepLens. She came up with ideas such as using DeepLens to detect an elderly person's fall off the bed or in coma state on the floor (we are planning to achieve this idea in the next creation project). I'm so proud she presented DeepLens and the idea to her middle school science class. It attracts a lot of attention and questions from her fellow middle schoolers.

What we learned

DeepLens and AWS AI ecosystem are very powerful to build solutions to help human better life. Amazon is on the right path of leading the industry to achieve the AI evolution. We are inspired and we are glad to have the opportunity of experiencing such an amazing product. 

What's next

Actually, the use case we solved in this project can be generalized as below: an intelligent system to detect the event, capture the image of the event, send real-time notification with accurate location for immediate response.

The next of our creation, we want to build a system to help seniors living alone. We want to use Sagemaker to train a deep learning model to detect person fall on the floor event and deploy this model to DeepLens -- to automatically detect elder person fall on floor or in coma state, and send real-time notification with accurate location of the fall event to urgent care team or relatives for fast reaction of rescue or assistant. 

Built with


Try it out

GitHub repo