CODEWITHSUNDEEP
conv-neural-networkyolo
optimal substructure
optimal substructure

Reputation: 125

How is output encoded for YOLO

If multiple boxes in a grid have same object lets say car, then for all the boxes, that has the car are bx, by, bh, bw are the same? Here bx, by are the center pixel of the car and bh, bw are the height and width of the bounding box.

Upvotes: 0

Views: 492

Answers (1)

Ankish Bansal
Ankish Bansal

Reputation: 1902

I also had the same confusion. Let me give you intuition behind this, how it deal with this problem.

In yolo, we have three important parameters IOU, class-confidence-score and box-confidence-score, which decide, which grid has more probability in all aspect and prune other lower probable grids.

So, even if there are many grids with the same bounding box dimensions, but the probability of object in the grid cell is varying, which change the box confidence score for the grids, which is calculated as pr(object).IOU. This way, yolo remove those grids.

Here is link, which have detained explanation of yolo.

Upvotes: 1

Related Questions