April 2018
Intermediate to advanced
334 pages
10h 18m
English
YOLO learns the class probabilities and the size of bounding boxes by performing regression and thus performs object detection on the input image. YOLO divides an image into SxS grids and each grid predicts N bounding boxes and confidence. This confidence value quantifies the accuracy of the bounding box and the occurrence of the object in the bounding box.
YOLO also predicts the class score of each box for all the classes in training. Thus, the summation of class scores over all boxes in the image also helps in calculating the class probability of the whole image and thus helping to predict the object. Since an image is divided in SxS grids and for each output N bounding boxes are predicted, therefore, SxSxN boxes are ...