SSD阅读笔记

论文下载地址:https://arxiv.org/abs/1512.02325
论文代码:https://github.com/weiliu89/caffe/tree/ssd


The Single Shot Detector (SSD)


创新点:

在5个feature map 上同时预测。



SSD结构



损失函数

N is the number of matched default boxes.
The confidence loss is the softmax loss over multiple classes confidences (c)
the weight term α is set to 1 by cross validation

你可能感兴趣的:(SSD阅读笔记)