YOLO algorithm - how it works?

vision
1

Hi.
I’m trying to learn about this algorithm, but any tutorial that i’ve read is like “YOLO is dividing image on SxS grid, where each cell is responsible for detecting an object…” it explains nothing to me.
How this algorithm is predicting boundry boxes which are larger than one grid cell?
Does it concatenate somehow those grid cells that are adjacent and predicting same class object?

I’ll be thankfull for good learning resources :slight_smile:

2

Have you had a look at the papers?
Yolov3
Yolov4

3

I’ll try to read this, but i have no experience in reading papers so it may be impossible to understand this :smiley:

4

I have read a tutorial [here fr example] to implement yolov3 from scratch, it even has a working code

In complement of reading the original papers, the idea is just to search the web for a good tutorial with working code explained step by step. There must be plenty of them available online now.

1 Like