Yolov5-7.0图像分类算法修改Resnet18/50主干网络流程
网上大多数都是基于yolov5算法的目标检测网络进行修改主干网络，我最近在尝试图像分类算法，流程如下：
以resnet50为例
1、打开models下的common.py文件，添加下面的代码：
'''
模型：resnet50
'''
class resnet501(nn.Module):
def __init__(self, ignore) -> None:
super().__init__()
model = models.resnet50(pretrained=True)
modules = list(model.children())
modules = modules[:6]
self.model = nn.Sequential(*modules)
def forward(self, x):
return self.model(x)
class resnet502(nn.Module):
def __init__(self, ignore) -> None:
super().__init__()
model = models.resnet50(pretrained=True)
modules = list(model.children())
modules = modules[6]
self.model = nn.Sequential(*modules)
def forward(self, x):
return self.model(x)
class resnet503(nn.Module):
def __init__(self, ignore) -> None:
super().__init__()
model = models.resnet50(pretrained=True)
modules = list(model.children())
modules = modules[7]
self.model = nn.Sequential(*modules)
def forward(self, x):
return self. Model(x)
2、打开yolo.py文件，添加
# 这行代码下添加
elif m is Expand:
c2 = ch[f] // args[0] ** 2
#新添加
elif m is resnet501 or m is resnet502 or m is resnet503:
c2 = args[0]3、新建resnet50.yaml文件，放在models文件夹下
# 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, resnet501, [512]], # 0
[-1, 1, resnet502, [1024]], # 1
[-1, 1, resnet503, [2048]], # 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)
]
4、测试yolo.py文件
在if __name__ == "__main__":下修改这行代码，添加resnet50.yaml文件
parser.add_argument('--cfg', type=str, default=ROOT/'resnet50.yaml', help='model.yaml')
运行：结果如下就是成功了
如果不是，检查以上步骤，哪里出错了。。。。
5、修改classify文件夹下的train.py文件，首先修改在parse_opt(known = False)函数下，添加加修改如下：
parser.add_argument('--model', type=str, default='resnet50', help='根路径下的预训练模型')
parser.add_argument('--cfg', type=str, default=ROOT/'models/resnet50.yaml', help='根路径下的预训练模型')然后，crtl+F定位 torchvision.models.__dict__ 这句代码，修改如下：
初始是这样的：
model = torchvision.models.__dict__[opt.model](weights='IMAGENET1K_V1' if pretrained else None)
修改之后：
model = torchvision.models.__dict__[opt.model](pretrained=opt.pretrained)再次定位 ClassificationModel ，修改如下
初始是这样的：
model = ClassificationModel(model=model, nc=nc, cutoff=opt.cutoff or 10) # convert to classification model
修改之后：
model = ClassificationModel(cfg = opt.cfg, model=model, nc=nc, cutoff=opt.cutoff or 10) # convert to classification model6、运行测试train.py文件，如下结果就是成功了
