基于轻量级yolov5实践开发家禽(鸭子)检测

之前做过生猪的检测识别，最近又有于淼、鸭子、鸡等等的检测，感觉AI应用到智慧农业方面还是一个有可为的方向的，这里主要是基于yolov5来开发实践一个家禽检测系统，首先看下效果：

鸭子检测识别：

鸭头计数效果：

接下来看下数据集：

YOLO格式标注文件如下：

实例标注数据如下：

0 0.0656 0.3222 0.0813 0.1556 
0 0.1339 0.2546 0.0615 0.1463 
0 0.2036 0.2676 0.1156 0.1352 
0 0.1797 0.3648 0.1281 0.1481 
0 0.3307 0.2898 0.0844 0.1722 
0 0.3036 0.412 0.101 0.1907 
0 0.3979 0.3787 0.1208 0.2056 
0 0.4865 0.3019 0.0625 0.1926 
0 0.5276 0.3704 0.0802 0.2481 
0 0.562 0.1981 0.0781 0.1481 
0 0.6078 0.2028 0.049 0.1204 
0 0.7052 0.2472 0.0875 0.1574 
0 0.6839 0.313 0.1094 0.1333 
0 0.7891 0.3083 0.1302 0.2241 
0 0.7344 0.3639 0.1542 0.2167 
0 0.5839 0.3306 0.0844 0.1241 
0 0.601 0.187 0.0604 0.1148 
0 0.6411 0.2176 0.051 0.1019 
0 0.7292 0.1741 0.0583 0.0815 
0 0.5057 0.1333 0.0385 0.1259 
0 0.5578 0.1102 0.0573 0.0648 
0 0.5307 0.0444 0.0344 0.063 
0 0.6188 0.0852 0.025 0.0963 

VOC格式标注文件如下所示：

实例标注内容如下所示：

<annotation>
    <folder>DATASET</folder>
    <filename>images/00010.jpg</filename>
    <source>
        <database>The DATASET Database</database>
        <annotation>DATASET</annotation>
        <image>DATASET</image>
    </source>
    <owner>
        <name>YSHC</name>
    </owner>    
    <size>
        <width>960</width>
        <height>540</height>
        <depth>3</depth>
    </size>
    <segmented>0</segmented>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>32</xmin>
            <ymin>127</ymin>
            <xmax>104</xmax>
            <ymax>216</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>104</xmin>
            <ymin>89</ymin>
            <xmax>156</xmax>
            <ymax>164</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>139</xmin>
            <ymin>107</ymin>
            <xmax>257</xmax>
            <ymax>177</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>115</xmin>
            <ymin>163</ymin>
            <xmax>240</xmax>
            <ymax>237</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>280</xmin>
            <ymin>98</ymin>
            <xmax>339</xmax>
            <ymax>195</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>254</xmin>
            <ymin>169</ymin>
            <xmax>339</xmax>
            <ymax>262</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>330</xmin>
            <ymin>142</ymin>
            <xmax>443</xmax>
            <ymax>259</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>436</xmin>
            <ymin>105</ymin>
            <xmax>498</xmax>
            <ymax>209</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>485</xmin>
            <ymin>121</ymin>
            <xmax>547</xmax>
            <ymax>273</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>504</xmin>
            <ymin>56</ymin>
            <xmax>571</xmax>
            <ymax>125</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>553</xmin>
            <ymin>72</ymin>
            <xmax>612</xmax>
            <ymax>135</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>577</xmin>
            <ymin>83</ymin>
            <xmax>645</xmax>
            <ymax>173</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>633</xmin>
            <ymin>77</ymin>
            <xmax>710</xmax>
            <ymax>167</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>697</xmin>
            <ymin>101</ymin>
            <xmax>828</xmax>
            <ymax>221</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>634</xmin>
            <ymin>125</ymin>
            <xmax>791</xmax>
            <ymax>248</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>448</xmin>
            <ymin>79</ymin>
            <xmax>487</xmax>
            <ymax>124</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>606</xmin>
            <ymin>110</ymin>
            <xmax>717</xmax>
            <ymax>175</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>517</xmin>
            <ymin>138</ymin>
            <xmax>600</xmax>
            <ymax>204</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>680</xmin>
            <ymin>73</ymin>
            <xmax>734</xmax>
            <ymax>111</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>466</xmin>
            <ymin>38</ymin>
            <xmax>507</xmax>
            <ymax>96</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>510</xmin>
            <ymin>39</ymin>
            <xmax>565</xmax>
            <ymax>71</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>577</xmin>
            <ymin>14</ymin>
            <xmax>608</xmax>
            <ymax>68</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>duck</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>495</xmin>
            <ymin>1</ymin>
            <xmax>533</xmax>
            <ymax>37</ymax>
        </bndbox>
    </object>
    
</annotation>

这里我一共做了两组实验从前面的效果图也不难看出的，第一组主要是检测鸭子，第二组主要是检测鸭头，在同样的参数配置下，我们看下结果对比。

【第一组：检测鸭子】

【第二组：检测鸭头】

从结果评估上面直观来看：鸭子检测的效果要略优于鸭头检测，这个可能是因为鸭头本身属于小目标的缘故吧。

模型方面使用的都是最轻量级的n系列的模型，如下：

#Parameters
nc: 1  # number of classes
depth_multiple: 0.33  # model depth multiple
width_multiple: 0.25  # layer channel multiple
anchors:
  - [10,13, 16,30, 33,23]  # P3/8
  - [30,61, 62,45, 59,119]  # P4/16
  - [116,90, 156,198, 373,326]  # P5/32


#Backbone
backbone:
  # [from, number, module, args]
  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
   [-1, 3, C3, [128]],
   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
   [-1, 6, C3, [256]],
   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
   [-1, 9, C3, [512]],
   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
   [-1, 3, C3, [1024]],
   [-1, 1, SPPF, [1024, 5]],  # 9
  ]


#Head
head:
  [[-1, 1, Conv, [512, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
   [-1, 3, C3, [512, False]],  # 13

   [-1, 1, Conv, [256, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)

   [-1, 1, Conv, [256, 3, 2]],
   [[-1, 14], 1, Concat, [1]],  # cat head P4
   [-1, 3, C3, [512, False]],  # 20 (P4/16-medium)

   [-1, 1, Conv, [512, 3, 2]],
   [[-1, 10], 1, Concat, [1]],  # cat head P5
   [-1, 3, C3, [1024, False]],  # 23 (P5/32-large)

   [[17, 20, 23], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
  ]

其余所有的参数设置均保持一致。