Question

When training the YOLO network, the center of the object should always fall in one of the SxS grid cells.​

Answer 1

Answer:

Compared to the other reason proposal classification network (fast RCNN) which performs detection on various reason proposals and the end up performing prediction multiple times of various reason in a image , Yolo architecture is more like FCNN (fully convolutional neural network) and passes the image (nxn) one's through the FCNN and output is mxm prediction

Explanation:

• Speed (45 frames per second — better than realtime)
• Network understand generalized object representation ( this allowed them to train the network one real world images and prediction on artwork was still fairly accurate).
• faster version (the smaller architecture)— 155 frames per sec but less accurate.