YOLOv5改进之替换Backbone为efficientnet b1

简介

BestYOLO：https://github.com/WangRongsheng/BestYOLO

BestYOLO是一个以科研和竞赛为导向的最好的YOLO实践框架!

目前BestYOLO是一个完全基于YOLOv5 v7.0 进行改进的开源库，该库将始终秉持以落地应用为导向，以轻便化使用为宗旨，简化各种模块的改进。目前已经集成了基于torchvision.models 模型为Backbone的YOLOv5目标检测算法，同时也将逐渐开源更多YOLOv5应用程序。

替换为efficientnet b1模型

修改common.py

在最后添加：

from torchvision import models
'''
模型：efficientnet_b1
'''
class efficientnet_b11(nn.Module):
    def __init__(self, ignore) -> None:
        super().__init__()
        model = models.efficientnet_b1()
        modules = list(model.children())
        modules = modules[0][:4]
        self.model = nn.Sequential(*modules)
    def forward(self, x):
        return self.model(x)
    
class efficientnet_b12(nn.Module):
    def __init__(self, ignore) -> None:
        super().__init__()
        model = models.efficientnet_b1()
        modules = list(model.children())
        modules = modules[0][4:6]
        self.model = nn.Sequential(*modules)
    def forward(self, x):
        return self.model(x)
    
class efficientnet_b13(nn.Module):
    def __init__(self, ignore) -> None:
        super().__init__()
        model = models.efficientnet_b0()
        modules = list(model.children())
        modules = modules[0][6:]
        self.model = nn.Sequential(*modules)
    def forward(self, x):
        return self.model(x)

如果不需要开启预训练权重，删除pretrained=True即可。

修改yolo.py

在elif m is Expand:下面添加：

elif m is efficientnet_b11 or m is efficientnet_b12 or m is efficientnet_b13:
	c2 = args[0]

修改.yaml配置

# YOLOv5 🚀 by Ultralytics, GPL-3.0 license

# Parameters
nc: 2  # 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

# YOLOv5 v6.0 backbone
backbone:
  # [from, number, module, args]
  [[-1, 1, efficientnet_b11, [40]],  # 0
   [-1, 1, efficientnet_b12, [112]],  # 1
   [-1, 1, efficientnet_b13, [1280]],  # 2
   [-1, 1, SPPF, [1024, 5]],  # 3
  ]

# YOLOv5 v6.0 head
head:
  [[-1, 1, Conv, [512, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 1], 1, Concat, [1]],  # cat backbone P4
   [-1, 3, C3, [512, False]],  # 7

   [-1, 1, Conv, [256, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 0], 1, Concat, [1]],  # cat backbone P3
   [-1, 3, C3, [256, False]],  # 11 (P3/8-small)

   [-1, 1, Conv, [256, 3, 2]],
   [[-1, 7], 1, Concat, [1]],  # cat head P4
   [-1, 3, C3, [512, False]],  # 14 (P4/16-medium)

   [-1, 1, Conv, [512, 3, 2]],
   [[-1, 3], 1, Concat, [1]],  # cat head P5
   [-1, 3, C3, [1024, False]],  # 17 (P5/32-large)

   [[11, 14, 17], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
  ]

.yaml配置文件中的depth_multiple和width_multiple可以同时设置为1.0试试，说不定会有不错的效果。

具体指标

models	layers	parameters	model size(MB)
efficientnet_b1	539	6595615	13.8