This website knows when you're angry!
Written by Harry Fairhead   
Saturday, 09 July 2011

A new mood detecting facility adds an extra dimension to an exciting facial recognition API - it can tell a smile from a sad look. It's easy to use, free and just waiting for you to think up ways of using it.

If you haven't heard of it is well worth knowing about. The website launched last year (2010) and now has over 20,000 sign-ups to its API. The idea is very simple: the API provides the basic facilities of face recognition for you to build into your own apps. It will already detect a face, tell you the sex of the person and if they are wearing glasses.

The latest development is mood recognition. You provide a photo and the API will return a data structure (json or xml) that contains lots of information about the faces found - eye positions, nose position, etc - and now a mood property has been added giving the classification - happy, sad, angry, etc - and the confidence level.  In addition you can also determine the state of the subject's lips - sealed, parted or kissing.

What this means is that you can now determine exactly what that facial expression means. Is your other half pleased with you, or is that just anger you see?

Could this be a computer equivalent of Dr Cal Lightman in TV's  "Lie to Me" ? 

It could well be in the future, but at the moment the range of moods recognized is fairly limited: happy/sad/angry/surprised/neutral. However for many applications these are sufficient. Also at the moment the system only works with still images but it would be easy to build an app that extracted a still from a video stream and submitted it to the API. Is your customer happy or angry?



Tim Roth as Dr Cal Lightman is in agreement with the anger recognized by (The captions and dotted lines were in the original, click to read code.)


 At the moment access to the API is free and you can upload and process 5000 photos per hour. The API is currently in beta and the free service will obviously change in the future as a paid service is rolled out. Even so the current allocation is very generous and would enable you to create and really test an app.

The API is easy to use and includes a face detection, recognition and tagging service. You can train the system to recognize a person by showing it a few photos and there is a Facebook interface which make tagging photos particularly easy.  The API is REST-based but there are libraries that allow you to work in JavaScript and PHP. Non-official libraries are available in Python, C#, Flash, Java and Ruby.There is also a sandboxed web app that lets you try out the API by selecting options and you can even specify a URL of a photo you would like to process or upload one.


This represents are real opportunity to invent some novel apps or to add face recognition to a bigger app. If you want some ideas have a look at the website . For example, Celebrityfind searches for pictures of celebrities and their look-alikes on Twitter. However uses go beyond the obvious and you can use the location of facial features such as eyes, mouth etc. to further process the image - cartoonize the person for example. This is too much fun to miss!

More Information


If you would like to be informed about new articles on I Programmer follow us on Twitter or Facebook or subscribe to our weekly newsletter.



Grow with Google Launches Generative AI Course

Grow with Google, in collaboration with MIT RAISE (Responsible AI for Social Empowerment and Education), is launching a no-cost Generative AI for Educators course.

Gemini 1.5 Pro Now Available

Google has released Gemini 1.5 Pro with improvements including Native Audio Understanding, System Instructions, and a JSON mode.

More News

Last Updated ( Saturday, 09 July 2011 )