Let's first understand the concepts regarding classification, localization, detection, and object detection problems, how they can be transformed into supervised machine learning problems, and subsequently, how they can be solved using a deep convolution neural network.

Refer to the following diagram:

Here's what we can infer:

• In the image classification problem, there is generally a (big) central object in an image and we have to recognize the object by assigning a correct label to the image. 
• Image classification with localization aims to find the location of an object in an image by not only assigning a label (class) the image (for example, a binary classification problem: whether there is a car in an image or not), but also finding a bounding box around the object, if there is one to be found.
• Detection goes a level further by aiming to identify multiple instances of same/different types of objects, by localizing all the objects (all instances) by marking their locations (the localization problem usually tries to find a single object location).
• The localization problem can be converted to a supervised machine learning multi-class classification and a regression problem in the following way: in addition to the class labels of the objects to be identified (with classification), the output vector corresponding to an input training image must also contain the location (bounding box coordinates relative to image size, with regression) of the object.
• A typical output data vector will contain eight entries for a 4-class classification. As shown in the following diagram, the first entry will correspond to whether or not an object of any from the three classes of objects (except the background). If one is present in an image, the next four entries will define the bounding box containing the object, followed by three binary values for the three class labels indicating the class of the object. If none of the objects are present, the first entry will be 0 and the others will be ignored: