[Python知识库]【魔改YOLOv5-6.x（下）】：YOLOv5+Ghostnet+BiFPN+CA

前言

【魔改YOLOv5-6.x（上）】：结合轻量化网络Shufflenetv2、Mobilenetv3和Ghostnet

【魔改YOLOv5-6.x（中）】：加入ACON激活函数、CBAM和CA注意力机制、加权双向特征金字塔BiFPN

本文使用的YOLOv5版本为v6.1，对YOLOv5-6.x网络结构还不熟悉的同学，可以移步至：【YOLOv5-6.x】网络模型&源码解析

训练设置：

$ python train.py --weights  --cfg yolov5s.yaml --data data/VOC2007.yaml -- hyp data/hyps/hyp.scratch-high.yaml --epochs 200 --device 0

• 实验环境为1个GTX 1080 GPU
• 数据集为VOC2007
• 超参数为hyp.scratch-low.yaml
• 训练200个epoch
• 其他参数均为源码中默认设置的数值

测试设置：

$ python val.py --weights yolov5s.pt --data data/VOC2007.yaml --img 832 --augment --half --device 0

• 使用val.py进行测试
• 使用TTA测试方法（augment）

?

模型组合

根据实验结果，作出如下选择：

• 使用Ghost模块，替换Backbone和Neck中的Conv模块与C3模块
• Backbone最后（SPPF之前）添加CA注意力机制
• Neck部分中添加一条BiFPN

具体修改模型的yaml文件如下：

Ghost+BiFPN

# YOLOv5 🚀 by Ultralytics, GPL-3.0 license

# Parameters
nc: 20  # 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

# YOLOv5 v6.0 backbone
backbone:
  # [from, number, module, args]
  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
   [-1, 1, GhostConv, [128, 3, 2]],  # 1-P2/4
   [-1, 3, C3Ghost, [128]],
   [-1, 1, GhostConv, [256, 3, 2]],  # 3-P3/8
   [-1, 6, C3Ghost, [256]],
   [-1, 1, GhostConv, [512, 3, 2]],  # 5-P4/16
   [-1, 9, C3Ghost, [512]],
   [-1, 1, GhostConv, [1024, 3, 2]],  # 7-P5/32
   [-1, 3, C3Ghost, [1024]],
   [-1, 1, SPPF, [1024, 5]],  # 9
  ]

# YOLOv5 v6.0 head
head:
  [[-1, 1, GhostConv, [512, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
   [-1, 3, C3Ghost, [512, False]],  # 13

   [-1, 1, GhostConv, [256, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
   [-1, 3, C3Ghost, [256, False]],  # 17 (P3/8-small)

   [-1, 1, GhostConv, [256, 3, 2]],
   [[-1, 14, 6], 1, Concat, [1]],  # cat head P4 <--- BiFPN change
   [-1, 3, C3Ghost, [512, False]],  # 20 (P4/16-medium)

   [-1, 1, GhostConv, [512, 3, 2]],
   [[-1, 10], 1, Concat, [1]],  # cat head P5
   [-1, 3, C3Ghost, [1024, False]],  # 23 (P5/32-large)

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

?

Ghost+CA

# YOLOv5 🚀 by Ultralytics, GPL-3.0 license

# Parameters
nc: 20  # 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

# YOLOv5 v6.0 backbone
backbone:
  # [from, number, module, args]
  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
   [-1, 1, GhostConv, [128, 3, 2]],  # 1-P2/4
   [-1, 3, C3Ghost, [128]],
   [-1, 1, GhostConv, [256, 3, 2]],  # 3-P3/8
   [-1, 6, C3Ghost, [256]],
   [-1, 1, GhostConv, [512, 3, 2]],  # 5-P4/16
   [-1, 9, C3Ghost, [512]],
   [-1, 1, GhostConv, [1024, 3, 2]],  # 7-P5/32
   [-1, 3, C3Ghost, [1024]],
   [-1, 1, CABlock, [1024, 32]],  # 9 CA <-- Coordinate Attention [out_channel, reduction]
   [-1, 1, SPPF, [1024, 5]],  # 10
  ]

# YOLOv5 v6.0 head
head:
  [[-1, 1, GhostConv, [512, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
   [-1, 3, C3Ghost, [512, False]],  # 14

   [-1, 1, GhostConv, [256, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
   [-1, 3, C3Ghost, [256, False]],  # 18 (P3/8-small)

   [-1, 1, GhostConv, [256, 3, 2]],
   [[-1, 15], 1, Concat, [1]],  # cat head P4
   [-1, 3, C3Ghost, [512, False]],  # 21 (P4/16-medium)

   [-1, 1, GhostConv, [512, 3, 2]],
   [[-1, 11], 1, Concat, [1]],  # cat head P5
   [-1, 3, C3Ghost, [1024, False]],  # 24 (P5/32-large)

   [[18, 21, 24], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
  ]

?

Ghost+BiFPN+CA

# YOLOv5 🚀 by Ultralytics, GPL-3.0 license

# Parameters
nc: 20  # 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

# YOLOv5 v6.0 backbone
backbone:
  # [from, number, module, args]
  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
   [-1, 1, GhostConv, [128, 3, 2]],  # 1-P2/4
   [-1, 3, C3Ghost, [128]],
   [-1, 1, GhostConv, [256, 3, 2]],  # 3-P3/8
   [-1, 6, C3Ghost, [256]],
   [-1, 1, GhostConv, [512, 3, 2]],  # 5-P4/16
   [-1, 9, C3Ghost, [512]],
   [-1, 1, GhostConv, [1024, 3, 2]],  # 7-P5/32
   [-1, 3, C3Ghost, [1024]],
   [-1, 1, CABlock, [1024, 32]],  # 9 CA <-- Coordinate Attention [out_channel, reduction]
   [-1, 1, SPPF, [1024, 5]],  # 10
  ]

# YOLOv5 v6.0 head
head:
  [[-1, 1, GhostConv, [512, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
   [-1, 3, C3Ghost, [512, False]],  # 14

   [-1, 1, GhostConv, [256, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
   [-1, 3, C3Ghost, [256, False]],  # 18 (P3/8-small)

   [-1, 1, GhostConv, [256, 3, 2]],
   [[-1, 15, 6], 1, Concat, [1]],  # cat head P4 <--- BiFPN change
   [-1, 3, C3Ghost, [512, False]],  # 21 (P4/16-medium)

   [-1, 1, GhostConv, [512, 3, 2]],
   [[-1, 11], 1, Concat, [1]],  # cat head P5
   [-1, 3, C3Ghost, [1024, False]],  # 24 (P5/32-large)

   [[18, 21, 24], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
  ]

?

实验记录

$ python val.py --weights yolov5s.pt --data data/VOC2007.yaml --img 832 --augment --half --device 0

?

结果分析

从实验结果来看，YOLOv5单独结合Ghostnet、CBAM和CA模块，效果都略有提升，但组合起来，提升效果并不明显，可能的原因是：

• 训练epoch太少，官方训练为300，后续可以尝试更多的epoch
• 超参数的选择，这里选择的是适配COCO2017数据集的hyp.scratch-low.yaml，对VOC2007数据集的训练会有一定的影响，后续可以直接训练COCO2017数据集
• CA模块添加的位置还可以更换（比如直接放到SPPF后面，或者放到Backbone中间的某些位置）

目前还没有想到更好的改进方法，欢迎大家前来交流，分享魔改YOLOv5的方法~