Inspiration

We wanted a way for Alexa to tell you if your clothes are appropriate for the weather outside before leaving home.

What we learned

Michael learned how to set up an Alexa Skill, he had not done it before. Zach learned how to use AWS Rekognition APIs. We both learned that training custom computer vision/deep learning models requires a lot of data, training time, and computing resources, more than we had available to us this weekend. We learned that using third-party APIs can make our lives much easier without sacrificing performance.

How we did it

In this project we set up a system with a Raspberry Pi, camera, and Alexa Dot. When the user says "Alexa, open Outfit Weather Checker", the Pi takes a picture and sends it to an AWS API called Rekognition. This API returns object detections (including clothing) and associated prediction confidences. The weather is returned from an OpenWeatherMap API using the user's IP address. Then, the sentence that Alexa speaks in response is formed, based on the weather and the clothing detected, recommending changes to the user's clothing that correspond to the weather outside.

Challenges

Connecting the Raspberry Pi to the Alexa using the Alexa Developer Console and a tunneling service called ngrok.

Built With

• alexa
• amazon-alexa
• amazon-web-services
• ngrok
• opencv
• python
• raspberry-pi
• rekognition