What is the difference between image classification, object detection, and localization?

These three sit on a ladder of difficulty, and mixing them up is one of the most common early mistakes. Here's the clean way to think about them.

• Classification: one image in, one label out ("this is a cat"). No idea where the cat is.
• Localization: one label plus one bounding box for the single, main object in the image.
• Detection: many objects at once, each with its own box and label ("3 cats, 1 dog, 2 cars").

There's a fourth rung people often want next: segmentation, which replaces the box with a pixel-perfect mask. If you only need a count or a rough location, detection is enough. If you need exact shape or area, you're into segmentation.

In practice, almost every real project I get asked about is detection, not classification. People rarely have one clean object per photo; they have a busy scene and want to know what's in it and where.