基于YOLO的酸枣病虫害检测识别实践

news2024/9/27 19:23:43

在我前面的博文中对于农作物病虫害的检测识别已经做过了，不过那个主要是针对水稻的，文章如下：

《基于yolov5的轻量级水稻虫害目标检测项目实践》

感兴趣的话可以自行移步阅读。

这里主要是针对酸枣常见的几种病虫害检测检测识别，首先看下效果：

接下来看下数据集概况：

YOLO格式标注数据文件如下：

实例标注内容如下：

1 0.25 0.34507 0.036364 0.030986
1 0.211364 0.460563 0.027273 0.033803
1 0.132955 0.239437 0.029545 0.04507
1 0.15 0.149296 0.036364 0.033803
1 0.177841 0.103521 0.053409 0.057746
1 0.179545 0.026761 0.025 0.033803
1 0.239205 0.05 0.014773 0.038028
1 0.331818 0.039437 0.013636 0.033803
1 0.411932 0.011972 0.0375 0.015493
1 0.677273 0.052817 0.034091 0.043662
1 0.786364 0.057746 0.086364 0.112676
1 0.877273 0.069718 0.045455 0.026761
1 0.095455 0.505634 0.079545 0.129577
1 0.150568 0.63169 0.0625 0.05493
1 0.201136 0.708451 0.056818 0.061972
1 0.180114 0.546479 0.017045 0.033803
1 0.224432 0.543662 0.014773 0.033803
1 0.140341 0.39507 0.021591 0.023944
1 0.204545 0.391549 0.015909 0.030986
1 0.238636 0.391549 0.015909 0.025352
1 0.247159 0.449296 0.019318 0.033803
1 0.265909 0.551408 0.011364 0.023944
1 0.280114 0.725352 0.017045 0.039437
1 0.339773 0.452817 0.015909 0.023944
1 0.389773 0.530282 0.034091 0.043662
1 0.340909 0.685915 0.022727 0.033803
1 0.451136 0.580282 0.015909 0.028169
1 0.430682 0.51831 0.018182 0.03662
1 0.449432 0.207042 0.019318 0.019718
1 0.516477 0.217606 0.019318 0.021127
1 0.531818 0.159155 0.011364 0.022535
1 0.589773 0.125352 0.020455 0.028169
1 0.622159 0.11338 0.023864 0.026761
1 0.684091 0.116197 0.018182 0.015493
1 0.555114 0.225352 0.014773 0.042254
1 0.597159 0.380282 0.017045 0.039437
1 0.549432 0.628873 0.0375 0.111268
1 0.480682 0.756338 0.052273 0.090141
1 0.440341 0.926761 0.071591 0.073239
1 0.589773 0.825352 0.095455 0.222535
1 0.74375 0.70493 0.132955 0.153521
1 0.665909 0.55 0.022727 0.057746
1 0.626705 0.571831 0.023864 0.033803
1 0.728409 0.411268 0.061364 0.169014
1 0.781818 0.296479 0.036364 0.074648
1 0.736932 0.182394 0.044318 0.043662
1 0.807386 0.15493 0.019318 0.030986
1 0.836932 0.174648 0.019318 0.028169
1 0.901705 0.247183 0.017045 0.038028
1 0.849432 0.294366 0.014773 0.076056
1 0.829545 0.421127 0.059091 0.059155
1 0.803409 0.482394 0.022727 0.035211
1 0.788636 0.556338 0.034091 0.08169
1 0.186364 0.222535 0.027273 0.109859
1 0.238636 0.249296 0.043182 0.033803

VOC格式标注数据文件如下所示：

实例标注内容如下所示：

<annotation>
    <folder>Data</folder>
    <filename>1f4707fe-6d90-4115-80fd-b0e5e3d4b18f.jpg</filename>.
    <source>
        <database>The Database</database>
        <annotation>Data</annotation>
        <image>Data</image>
    </source>
    <owner>
        <name>YSHC</name>
    </owner>    
    <size>
        <width>495</width>
        <height>372</height>
        <depth>3</depth>
    </size>
    <segmented>0</segmented>
    
    <object>        
        <name>G</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>588</xmin>
            <ymin>314</ymin>
            <xmax>632</xmax>
            <ymax>368</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>G</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>784</xmin>
            <ymin>526</ymin>
            <xmax>807</xmax>
            <ymax>534</ymax>
        </bndbox>
    </object>
    
    <object>        
        <name>G</name>
        <pose>Unspecified</pose>
        <truncated>0</truncated>
        <difficult>0</difficult>
        <bndbox>
            <xmin>803</xmin>
            <ymin>557</ymin>
            <xmax>848</xmax>
            <ymax>578</ymax>
        </bndbox>
    </object>
    
</annotation>

这里我使用的是yolov5s系列的模型，yaml文件如下所示：

#Parameters
nc: 3  # number of classes
depth_multiple: 0.33  # model depth multiple
width_multiple: 0.50  # 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)
  ]

默认100epoch的迭代计算，结果详情如下：

混淆矩阵：