Mid point object?

[izav]

Hello,

I think I don’t fully understand how object detection works… In the course Andrew says that the mid point of the object is detected, how is that done? And also, if an object is bigger than the grid cell size, how does the net know that those are parts of the same object?

Thank you in advance!

[Dyxuki]

Hey, I got the same question while learning this, maybe this thread can help you

Questions about YOLO

Hello, so i’m going through C4W3 courses (object detection with CNN), and I still cannot get a good enough intruition on why algos such as YOLO works. If I understand correctly, we do not use formal sliding windows because they are very computational expensive, so we use an alternative solution which deals with convolution layers. So when it comes to the YOLO algo for example, we used grid in split the input image, so the y_hat at the end has a the same height and width than the grid. It loo…

[izav]

That was very useful! Thanks a lot

However, I still don’t understand how the “mid-point” of the object is defined… When we train the network, we don’t input such “mid-point”, so how is the network able to determine if the mid-point is in the cell or not?

[TMosh]

The mid-points are assigned in creating the training set. Someone looks at each image, clicks a mouse on what they think is the center of the object, and draws a box around it.

The system learns that when it detects a specific pattern of pixels, it knows to output a box of the size it has learned for similar objects in the training set.