Understanding the testing metrics
In this section, we will cover the testing metrics. We will look at the two matrices that will help us understand how to test the object detection application. These testing matrices are as follows:
Intersection over Union (IoU)
mean Average Precision (mAP)
Intersection over Union (IoU)
For detection, IoU is used in order to find out whether the object proposal is right or not. This is a regular way to determine whether object detection is done perfectly or not. IoU generally takes the set, A, of proposed object pixels and the set of true object pixels, B, and calculates IoU based on the following formula:
Generally, IoU >0.5, which means that it was a hit or that it identified the object pixels or boundary box for the object; otherwise, it fails. This is a more formal understanding of the IoU. Now, let's look at the intuition and the meaning behind it. Let's take an image as reference to help us understand the intuition behind this matrix. You can refer...
