I have question related to yolo box to corners. I did not understand this concept. What does Yolo box to corners describes about? Does it refers to adjusting input image ? Why do we have to represent box by corners rather than x, y, h and w co-ordinates ? Why do we need to represent the boxes by corners ?
This is discussed in the text of the notebook. See the section on the iou function. If you are representing a rectangular box overlaid on the image space, you have several choices for how to specify that. In some of the algorithms here, it’s easier to use the centroid of the box and the height and width. For the purposes of iou, it’s way easier to know the coordinates of the upper left and lower right corners, so we need to convert between those two representations.