• 暴力目標檢測                                                                                                 

  •  滑動視窗（從左到右，從上到下）（多個尺寸大小）                               

•   輸入圖片歸一化                                                                                                                      

• R-CNN                                                                                                            

  •  選擇性搜尋                                                                                            

•  ROI圖片歸一化                                                                                     
•  使用選擇搜尋有風險，合併太大，會損失目標（損失的是重疊目標）。合併太小，資料太多                                                                     

總：

• Fast R-CNN                                                                                                     

•  邊界框迴歸器                                                                                         

•  在卷積特徵提取後使用Selectiv Search搜尋方法                                     

•  特徵圖塊歸一化（ROI池化）                                                                

•  Faster R-CNN                                                                 

•  候選區域網路（RPN）                                                                           

• YOLO                                       

  •  訓練迴歸框                      
  •  暴力設S*S的網路           
    •  太大不精確               
    •  太小分類不好做        

•  SSD                                                          

•  多尺度檢測                             

•  卷積測物，加框先驗