YOLOv7轻量化：ghost

玩大侠

已于 2023-12-12 03:32:34 修改

阅读量302

点赞数

分类专栏：轻量化文章标签： YOLO 前端

于 2023-12-07 22:41:31 首次发布

本文链接：https://blog.csdn.net/qq_43409396/article/details/134866962

版权

轻量化专栏收录该内容

4 篇文章 0 订阅

订阅专栏

本文深入探讨了YOLOv7在轻量化方面的改进，特别是引入了Ghost模块，该模块通过简单操作极大地提高了网络的效率和性能。通过对骨干网路由的优化，Ghost模块成为提升前端推理速度的关键技术。

摘要由CSDN通过智能技术生成

骨干网路由ghost网络替代

# parameters
nc: 80  # number of classes
depth_multiple: 1.0  # model depth multiple
width_multiple: 1.0  # layer channel multiple

# anchors
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

# yolov7-tiny backbone
backbone:
  # [from, number, module, args] c2, k=1, s=1, p=None, g=1, act=True
  [[-1, 1, Conv, [16, 3, 2, 1]],            # 0-P1/2  ch_out, kernel, stride, padding, groups
   [-1, 1, G_bneck, [16, 16, 3, 1]],        # 1  ch_out, ch_mid, dw-kernel, stride

   [-1, 1, G_bneck, [24, 48, 3, 2]],        # 2-P2/4
   [-1, 1, G_bneck, [24, 72, 3, 1]],        # 3

   [-1, 1, G_bneck, [40, 72, 3, 2, True]],  # 4-P3/8
   [-1, 1, G_bneck, [40, 120, 3, 1, True]], # 5

   [-1, 1, G_bneck, [80, 240, 3, 2]],        # 6-P4/16
   [-1, 3, G_bneck, [80, 184, 3, 1]],        # 7
   [-1, 1, G_bneck, [112, 480, 3, 1, True]],
   [-1, 1, G_bneck, [112, 480, 3, 1, True]],

   [-1, 1, G_bneck, [160, 672, 3, 2, True]], # 10-P5/32
   [-1, 1, G_bneck, [160, 960, 3, 1]],       # 11
   [-1, 1, G_bneck, [160, 960, 3, 1, True]],
   [-1, 1, G_bneck, [160, 960, 3, 1]],
   [-1, 1, G_bneck, [160, 960, 3, 1, True]],
   [-1, 1, Conv, [960]],
  ]


# yolov7-tiny head
head:
  [[-1, 1, Conv, [256, 1, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-2, 1, Conv, [256, 1, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-1, 1, SP, [5]],
   [-2, 1, SP, [9]],
   [-3, 1, SP, [13]],
   [[-1, -2, -3, -4], 1, Concat, [1]],
   [-1, 1, Conv, [256, 1, 1, None, 1, nn.LeakyReLU(0.1)]],
   [[-1, -7], 1, Concat, [1]],
   [-1, 1, Conv, [256, 1, 1, None, 1, nn.LeakyReLU(0.1)]],  # 37 24
  
   [-1, 1, Conv, [128, 1, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [9, 1, Conv, [128, 1, 1, None, 1, nn.LeakyReLU(0.1)]], # route backbone P4
   [[-1, -2], 1, Concat, [1]],
   
   [-1, 1, Conv, [64, 1, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-2, 1, Conv, [64, 1, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-1, 1, Conv, [64, 3, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-1, 1, Conv, [64, 3, 1, None, 1, nn.LeakyReLU(0.1)]],
   [[-1, -2, -3, -4], 1, Concat, [1]],
   [-1, 1, Conv, [128, 1, 1, None, 1, nn.LeakyReLU(0.1)]],  # 47 34
  
   [-1, 1, Conv, [64, 1, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [5, 1, Conv, [64, 1, 1, None, 1, nn.LeakyReLU(0.1)]], # route backbone P3
   [[-1, -2], 1, Concat, [1]],
   
   [-1, 1, Conv, [32, 1, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-2, 1, Conv, [32, 1, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-1, 1, Conv, [32, 3, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-1, 1, Conv, [32, 3, 1, None, 1, nn.LeakyReLU(0.1)]],
   [[-1, -2, -3, -4], 1, Concat, [1]],
   [-1, 1, Conv, [64, 1, 1, None, 1, nn.LeakyReLU(0.1)]],  # 57 44
   
   [-1, 1, Conv, [128, 3, 2, None, 1, nn.LeakyReLU(0.1)]],
   [[-1, 34], 1, Concat, [1]],
   
   [-1, 1, Conv, [64, 1, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-2, 1, Conv, [64, 1, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-1, 1, Conv, [64, 3, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-1, 1, Conv, [64, 3, 1, None, 1, nn.LeakyReLU(0.1)]],
   [[-1, -2, -3, -4], 1, Concat, [1]],
   [-1, 1, Conv, [128, 1, 1, None, 1, nn.LeakyReLU(0.1)]],  # 65 52
   
   [-1, 1, Conv, [256, 3, 2, None, 1, nn.LeakyReLU(0.1)]],
   [[-1, 24], 1, Concat, [1]],
   
   [-1, 1, Conv, [128, 1, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-2, 1, Conv, [128, 1, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-1, 1, Conv, [128, 3, 1, None, 1, nn.LeakyReLU(0.1)]],
   [-1, 1, Conv, [128, 3, 1, None, 1, nn.LeakyReLU(0.1)]],
   [[-1, -2, -3, -4], 1, Concat, [1]],
   [-1, 1, Conv, [256, 1, 1, None, 1, nn.LeakyReLU(0.1)]],  # 73 60
      
   [44, 1, Conv, [128, 3, 1, None, 1, nn.LeakyReLU(0.1)]],
   [52, 1, Conv, [256, 3, 1, None, 1, nn.LeakyReLU(0.1)]],
   [60, 1, Conv, [512, 3, 1, None, 1, nn.LeakyReLU(0.1)]],

   [[61,62,63], 1, IDetect, [nc, anchors]],   # Detect(P3, P4, P5)
  ]