YOLOV5增加检测头

以下内容均以yolov5的7.0版本为准

以下手绘图（忽略白色粗线）

以下配置文件yolov5s_anchors4.yaml 将下面内容复制到models文件下

# YOLOv5 🚀 by Ultralytics, GPL-3.0 license

# Parameters
nc: 80  # number of classes
depth_multiple: 0.33  # model depth multiple
width_multiple: 0.50  # layer channel multiple
anchors:
  - [4,5, 8,10, 22,18]
  - [10,13, 16,30, 33,23]  # P3/8
  - [30,61, 62,45, 59,119]  # P4/16
  - [116,90, 156,198, 373,326]  # P5/32


# 当前结构数量:max(round(number * depth_multiple), 1)(Conv数量   C3内部的残差数量)
# args :作为参数传入类中(内部元素可能会做调整)
# 1.Conv/C3/SPPF：输入维度：ch[-1] (ch 会储存上一个结构的输出维度)    输出维度：args[0] * width_multiple 
         # eg:args [64,6,2,2]  => [3,32,6,2,2]    Conv2d(3,32,6,2,2)  in,out,k,s,p   (若没有p,则为k//2)

# 2.Concat：输入维度 : c2 = sum(ch[x] for x in from)   args:[1] 表明 torch.cat(x, 1)在第一维上拼接

# YOLOv5 v6.0 backbone
backbone:
  # [from, number, module, args]  # (640,640) (为了方便查看 输入先不乘 width_multiple) 
  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2   (64,320,320)
   [-1, 1, Conv, [128, 3, 2]],    # 1-P2/4   (128,160,160)
   [-1, 3, C3, [128]],            # 2        (128,160,160)            
   [-1, 1, Conv, [256, 3, 2]],    # 3-P3/8   (256,80,80)
   [-1, 6, C3, [256]],            # 4        (256,80,80)           
   [-1, 1, Conv, [512, 3, 2]],    # 5-P4/16  (512,40,40)
   [-1, 9, C3, [512]],            # 6        (512,40,40)                  
   [-1, 1, Conv, [1024, 3, 2]],   # 7-P5/32  (1024,20,20)
   [-1, 3, C3, [1024]],           # 8        (1024,20,20)            
   [-1, 1, SPPF, [1024, 5]],      # 9        (1024,20,20)          
  ]

# YOLOv5 v6.0 head
head:
  [[-1, 1, Conv, [512, 1, 1]],                     # 10 (512,20,20)
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],     # 11 (512,40,40)
   [[-1, 6], 1, Concat, [1]],  # cat backbone P4     12 (1024,40,40)
   [-1, 3, C3, [512, False]],                      # 13 (512,40,40)                 csp2_x 无残差

   [-1, 1, Conv, [256, 1, 1]],                     # 14 (256,40,40)
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],     # 15 (256,80,80)
   [[-1, 4], 1, Concat, [1]],  # cat backbone P3     16 (512,80,80)




   [-1, 3, C3, [256, False]],                      # 17 (256,80,80)
   [-1, 1, Conv, [128, 1, 1]],                     # 18 (128,80,80)
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],     # 19 (128,160,160)
   [[-1, 2], 1, Concat, [1]],                      # 20 (256,160,160)

   [-1, 3, C3, [128, False]],                      # 21 (128,160,160)
   [-1, 1, Conv, [128, 3, 2]],                     # 22 (128,80,80)
   [[-1, 18], 1, Concat, [1]],                     # 23 (256,80,80)


   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)     24 (256,80,80)

   [-1, 1, Conv, [256, 3, 2]],                     # 25 (256,40,40)
   [[-1, 14], 1, Concat, [1]],  # cat head P4        26 (512,40,40)
   [-1, 3, C3, [512, False]],  # 20 (P4/16-medium)   27 (512,40,40)
   [-1, 1, Conv, [512, 3, 2]],                     # 28 (512,20,20) 
   [[-1, 10], 1, Concat, [1]],  # cat head P5      # 29 (1024,20,20)
   [-1, 3, C3, [1024, False]],  # 23 (P5/32-large) # 30 (1024,20,20)

   [[21, 24, 27,30], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
  ]

此时运行yolo.py，查看是否配置完成（别忘记改cfg路径），成功则配置完成，此时的yolov5(7)已经拥有了四个检测头