Chapter 4 · The deep learning era (2006–2016)2015
Object detection
Object detection finds each thing in an image and draws a labelled box around it.
Chapter 4 · The deep learning era (2006–2016)2015
Object detection finds each thing in an image and draws a labelled box around it.
A picture goes in…
A detector answers two questions at once: what is it, and where is it? The where is a box.
An image classifier only asks what is in a picture. Detection also asks where. For every object it gives a box and a label.
The picture holds a cat, a ball and a chair. Faint dashed boxes are guesses that were removed. Aim for one box per object and nothing else.
Every object has exactly one box. The boxes are hand-made for this lesson, so it is a simplified teaching version.
What does object detection give you that plain image classification does not?
You raise the confidence threshold a lot. What is most likely?
Three boxes sit on the same cat. What cleans this up?