Computer science from A to Z

F is for Form (Shape)

© INRIA Sophie Auvin - F comme Forme

Recognising a face, understanding a word you hear, distinguishing "spam" from legitimate e-mails, spotting an abnormal ECG: for a computer, all these things more or less come down to the same thing... shape recognition.

The aim is always to examine a set of data and place them in the right category. For the computer, the nature of the data matters little: images, sounds, texts or measurement results are simply series of numbers.

What counts is the way in which the categories are defined. Sometimes there is an explicit definition: “any message containing any of the following words will be rejected”. Often there is a less obvious indication which is interpreted in statistical terms: the ECG should have peaks that are fairly high but not too wide. Finally, some categories can only be represented by examples: recognising a face means searching through the photos available to you and finding the one that most resembles the face in question.

Our brain has an innate capacity to recognise certain shapes, but a great deal of inventiveness is required to give a machine such capabilities!

Did you know...?

© INRIA / Sophie Chauvin - Forme

In some shape recognition tasks, the categories are unknown: we must create them through a statistical analysis of the data.

Stock market experts can therefore optimise their investments by identifying the shares whose value is sensitive to the same factors.

Keywords: Computer science from A to Z Science awareness

Top