Google AIY Emotion Detection

A device to detect if you are sad and send dog pictures

This project started with assembling the Google AIY Vision Kit, this was very enjoyable and gave me a good insight into what makes up the device.

This code is a fork of the Google example face_camera_detection.

Building on top of this example, I created a counter, which allowed me to keep track of how consistently happy or sad the face in the frame is.

if avg_joy_score(faces) > 0.8:
	if joy_counter < 0:
		joy_counter = 0
 		joy_counter += 1
if avg_joy_score(faces) < 0.1:
 	if joy_counter > 0:
 		joy_counter = 0
    joy_counter -= 1
if joy_counter > 20:
joy_counter = 0

Sending Emails

The code for sending emails is nested under a conditional of if joy_counter < -20 in a similar way to the Happy branch is defined above.

For sending emails, I first had to work with the Reddit API, this is fairly simple as adding /random.json to the end of a subreddit will give you a response with a random post from that subreddit. So I just used urllib to make a request in the following way

request_url = urllib.request.urlopen("")

Then it comes to parsing the response, as the URL to the image is deep in the JSON response. The json package works for this, with result being set as a string of the URL

result = json.loads([0]["data"]["children"][0]["data"]["url"]

Then it comes to working with SendGrid, this is very simple with the sendgrid Python library

message = Mail(
	subject='Sending with Twilio SendGrid is Fun',
	html_content='<img src='+result+'>')
		sg = SendGridAPIClient(env.str('SENDGRID_API_KEY'))
		response = sg.send(message)
	except Exception as e:

With those simple modifications, it will now send the image when you get sad.