YoloV1:对比R-CNN
出处:
《You Only Look Once: Unified, Real-Time Object Detection》
背景:
RCNN存在region proposal 过程,Selective Search每张图花费1~2秒,不能做到real-time
Proposals太多会出现很多假阳例
思想:
提出一个简单的卷积神经网络同时预测边框和预测类别
使用整个图的特征来预测来减少误差
不需要预处理和后处理
结构:
24层卷积+2层FC,3*3卷积下采样,1*1卷积减少特征空间
Our system divides the input image into a 7* 7 grid. If the center of an object falls into a grid cell, that grid cell is responsible for detecting that object. Each grid cell predicts a bounding box and class probabilities associated with that bounding box
输入是448*448*3,输出是7*7*30, 7*7对应每个grid,而20代表20个类的概率和10