How does SSD ( single shot multibox )object detection model computes the the confidence score and the bounding box coordinates, referring to SSD paper, it said that this process is done with the aid of small convolutional Kernels. The kernels will output four real values (bbox offsets) and a vector with class scores. Can someone explain how this process is done with more clarification?
SSD: Single Shot MultiBox Detector available at:
Thanks in advance.