Face Recognizing Doorbell

Build a talking, face-recognizing doorbell for about $100

Recently, I set out to install a doorbell in my new house and thought: why doesn’t my doorbell tell me who is at the door?

Most of my DIY projects end up costing more than the equivalent product, even if I value my time at $0 per hour. Something about supply chains and economies of scale, I guess. (But I have way more fun making things myself.) In the course of this project, I built a door camera that is not only way cheaper than my Dropcam, but it has some genuinely useful features that, for some reason, aren’t available on the market yet.

August Lock keypad and face-recognizing Raspberry Pi
Figure 1. My front door with a doorbell, August Lock keypad and face-recognizing Raspberry Pi. Image by Lukas Biewald.

Here’s what we’re going to build: a $60 Raspberry Pi-powered security camera setup that takes pictures, posts them to the cloud, and then does face recognition. You could also stream the data to Amazon S3, making it a full-fledged Dropcam replacement. While Nest charges $100 a year for keeping the last 10 days of video footage, you could keep a year of camera footage in S3 for around $20. If you used Amazon Glacier, that cost would go down to around $4.Get O'Reilly's AI newsletter

Machine learning with Amazon Rekognition

This tutorial will focus on the machine learning part—using Amazon’s new Rekognition service to do face recognition on your guests, and send that to your Amazon Echo so you will always know who’s at your door. In order to build a reliable service, we’ll also make use of one of Amazon’s coolest and most useful products: Lambda.

Ingredients:

Amazon Echo Dot ($50)

Raspberry PI V3 ($38) (This project would also work with a Pi v2 and USB Wifi)

Raspberry PI-Compatible Camera ($16)

Raspberry Pi Case ($6)

16GB SD Card ($8)

Total: $118

We will use Amazon’s S3, Lambda, and Rekognition services for the face matching. These services are free to get started, and you can recognize thousands of people at your door every month for pennies.

Setting up the Raspberry Pi

If you’ve done any of my other Raspberry Pi tutorials, much of this will be familiar.

First, download Noobs from the Raspberry Pi Foundation and follow their setup instructions. This mainly involves copying Noobs onto an SD card and then plugging the SD card into your Pi. Then plug a mouse, keyboard, and monitor into your Pi and follow the setup instructions, which have gotten much more accessible since the launch of Pixel, the new desktop environment.

Raspberry Pi
Figure 2. Raspberry Pi on my desk with tiny monitor and keyboard. Image by Lukas Biewald.

Next, change the name of your Pi to something you can remember, so you can SSH into it. There’s good instructions for this on howtogeek—you need to modify the /etc/hosts and /etc/hostname files and give your Pi a name. I like to name all of my security camera Raspberry Pi after characters on my favorite TV show, “It’s Always Sunny in Philadelphia,” so I named my front door camera “Dennis.” That means I don’t need to remember an IP address, and I can SSH to dennis.local at all times, even if my router gets reset.

Next, you should attach your Raspberry Pi Camera to your PI. Remember, the tape should face the Ethernet jack—I’ve probably Googled that a hundred times by now. Note: you might want to buy a wide-angle camera for a bigger field of view; you also might want to buy an infrared camera to add night vision.

SAFARI

Learn faster. Dig deeper. See farther.

Join Safari. Get a free trial today and find answers on the fly, or master something new and useful.

Learn more 

Raspberry Pi with camera and case
Figure 3. Raspberry Pi with camera and case, ready to be installed. Image by Lukas Biewald.

You also probably want to put this whole contraption in a case to protect it from the weather. You’re also going to need to connect the Pi to power, through a micro USB cable. (I had already drilled a small hole to connect my Dropcam to an indoor power outlet, so I had a USB cable sitting in the right spot.)

I’ve actually installed a couple of these around the house at this point. The camera ribbon cable is so thin, you can potentially mount the Pi inside and slide the cable over the door like I did from my laboratory (garage).

Raspberry Pi camera
Figure 4. Raspberry Pi camera poking out from my garage door. Image by Lukas Biewald.

Next, you need to install RPi-Cam-Web-Interface. This is a really useful piece of software that serves up a continuous stream from the Pi camera over http. Follow the installation instructions and choose NGINX for your webserver. There is a really useful configuration file at /etc/raspimjpeg where you can change a ton of stuff.

There's more over at https://www.oreilly.com/ideas/build-a-talking-face-recognizing-doorbell-for-about-100

More News Stories

April 20, 2017
What We Can Learn From AI's Mistakes

What We Can Learn From AI's Mistakes

Read story
April 18, 2017
The Voice Controlled, Face Recognizing, Drone Journey

Building a drone

Read story