Single Shot Multibox Detector - myzwisc/CS766-Project GitHub Wiki

The SSD approach is based on a feed-forward convolutional network that produces a fixed-size collection of bounding boxes and scores for the presence of object class instances in those boxes, followed by a non-maximum suppression step to produce the final detections. There are basically two characteristics:

Speed: remove region proposal with default region; predict box and class in the mean time, in a single network instead of separate network like RPN;

Precision: use multi-scale feature maps for detection