This is a prototype Node.js and React app that captures an image or video stream from your built-in camera and analyzes it using Microsoft Azure's Emotion API. A history of emotion scores is plotted using Highcharts.
You'll need Node.js, and then you'll do some familiar things:
git clone https://github.com/mphonic/snap-to-emo.git
cd snap-to-emo
npm install
Having done that, you'll need to get an Azure account and subscription keys to the Emotion API. These are both available at free tiers.
After signing up, create a copy of /app/credentials/emotion-api-default.js in the credentials folder, and name it emotion-api.js. Copy the subscription key and URL you receive into emotion-api.js:
/app/credentials/emotion-api.js
module.exports = Object.freeze({
EMOTION_API_URL: 'your-subscription-url',
EMOTION_API_KEY: 'your-api-key'
});
Then, run
npm run dev
A tab should open in your web browser at http://localhost:3000/build. If it doesn't, enter the URL manually in your browser (or check your terminal for error messages).
The page should be pretty self-explanatory. Once you allow access to your camera, an interface appears. Take a decent picture featuring your face and get your emotion scores, which should appear in the knobs below, unless the API can't recognize a face in the image. You can adjust the knobs if you disagree with the algorithmic assessment, and then save the data into localStorage. A chart will appear showing all of your saved scores (since this app uses localStorage, don't expect to keep a long-term record -- this is only a proof-of-concept).
There is also an option to stream video to the Emotion API by going to http://localhost:3000/build/?stream=1. Be advised that this sends many requests to the API, which will be added to your Azure account and might be halted due to exceeding rate limits. There is a setTimeout in the CamStream class that can slow down the rate of requests.