Question of the Day

by JS

What is an object?

UPDATE: This is such a general question, I thought it might be worth providing a specific example. In the dominant paradigm of computer vision research, an object is an array of pixels with a particular label. The main research challenge is identifying some model that explains the variations in pixel arrays for each kind of label. As an example, if you have a thousand images of bikes, the actual values of the pixels in each image will vary quite a bit even though all the labels are the same. Using this kind of model, you’d like to be able to determine if an arbitrary array of pixels in some new image contains some kind of bike.

Now “labelled array of pixels” isn’t a very rich notion of objects, and it usually depends on having some population of already labelled pixel arrays around. Simplicity aside, having this kind of model would be incredibly useful for many applications.

Other disciplines where the notion of “object” comes up quite a bit have developed somewhat more comprehensive definitions. The challenge of today’s question is to come up with some common sense definition of your own.