How to solve this

Course Q&A Deep Learning Specialization Convolutional Neural Networks
1

16593373360701659414894283661885
165933733607016594148942836618851920×1080 103 KB

Doesn’t non max suppression chooses the bounding box with highest probability?What is the use of IOU here? IOU is used to evaluate.for that we need ratio of predicted and original.We are outputting the values of Pc.

Pls explain the solution stepwise. Thankyou in advance

2

IOU is the method for keeping both overlapping boxes if it is probable that they contain different objects, and keeping only one box if it is probable they contain the same object.

3

I think majority is covered by this thread.

The key point is, we sometimes forget the 2nd case below.

image
image1103×414 19 KB

In the 2nd case, even if two boxes are overlapping, we keep two boxes (do not select the highest one), since those can be two different objects. That’s controlled by IoU threshold value. In this quiz, it is 0.5. That’s the key point for this quiz, I think.