【超详细】基于YOLOv8改进1-Drone-YOLO复现

主要内容如下：

1、论文Drone-YOLO解析
2、网络结构
3、论文实验结果
4、论文结构复现，包括：主干网络添加RepVGG、增强小目标检测头、三明治结构

运行环境：Python=3.8（要求>=3.8），torch1.12.0+cu113（要求>=1.8）
复现源码：https://github.com/Bigtuo/Drone-YOLO

1 论文Drone-YOLO解析

论文：Drone-YOLO
文章改进点主要为 3 个部分：
改进 1：在网络的主干部分，我们使用 RepVGG 结构的重新参数化卷积模块作为下采样层。
该模块优势：在训练过程中，这种卷积结构同时训练 3×3 和 1×1 卷积。在推理过程中，两个卷积核被合并为一个 3×3 卷积层。这种机制使网络能够在不影响推理速度或扩大模型大小的情况下学习更稳健的特征。

改进 2：三明治结构（作者提出），相比于原先的 concat 操作，多融合了一个上层大尺度特征（深度可分离卷积下采样），文章说增强了多尺度检测头收集待检测对象的空间定位信息的能力。

改进 3：加了一个小目标分支 160*160（P2）【论文涨点核心】；

总网络结构：

三明治结构：

2 论文实验结果

实验结果 1：
如下图所示，在 yolov8-n 基础改进上，精度有所提升，参数变化不大，但是推理速度大幅增加（主要是三明治结构+P2 层的影响，REPVGG 不影响推理速度）。NVIDIA Tegra TX2：

实验结果 2（消融实验）：
如下图，改进 3（加小目标分支）是提升最显著，为 6.8 个点，无人机小目标较多（P2 是 v8 自带配置）；作者主要提出的三明治改进提升最小，为 0.2 个点（该改进类似 v6 的 BIC 操作，加入了深度可分离卷积）；Repvgg 是常用的涨点操作，训练时通过增加模型复杂度来提升模型效果，为 0.6 个点；

3 结构复现

注意：这里只复现 YOLOv8s 版本。

3.1 加小目标层

yolov8 自带 yolov8-p2.yaml 文件，如下：

3.2 主干网络下采样换成 RepVGG 模块

步骤 1：复制 1 份 P2.yaml 文件改名为 yolov8-p2-repvgg.yaml

步骤 2：在 block.py 中添加 RepVGGBlock 模块
（具体实现代码见 ultralytics/ultralytics/nn/modules/block.py 尾部）

# Ultralytics YOLO 🚀, AGPL-3.0 license
"""Block modules."""

import torch
import torch.nn as nn
import torch.nn.functional as F

from .conv import Conv, DWConv, GhostConv, LightConv, RepConv
from .transformer import TransformerBlock

__all__ = ('DFL', 'HGBlock', 'HGStem', 'SPP', 'SPPF', 'C1', 'C2', 'C3', 'C2f', 'C3x', 'C3TR', 'C3Ghost',
           'GhostBottleneck', 'Bottleneck', 'BottleneckCSP', 'Proto', 'RepC3', 'RepVGGBlock')


class DFL(nn.Module):
    """
    Integral module of Distribution Focal Loss (DFL).

    Proposed in Generalized Focal Loss https://ieeexplore.ieee.org/document/9792391
    """

    def __init__(self, c1=16):
        """Initialize a convolutional layer with a given number of input channels."""
        super().__init__()
        self.conv = nn.Conv2d(c1, 1, 1, bias=False).requires_grad_(False)
        x = torch.arange(c1, dtype=torch.float)
        self.conv.weight.data[:] = nn.Parameter(x.view(1, c1, 1, 1))
        self.c1 = c1

    def forward(self, x):
        """Applies a transformer layer on input tensor 'x' and returns a tensor."""
        b, c, a = x.shape  # batch, channels, anchors
        return self.conv(x.view(b, 4, self.c1, a).transpose(2, 1).softmax(1)).view(b, 4, a)
        # return self.conv(x.view(b, self.c1, 4, a).softmax(1)).view(b, 4, a)


class Proto(nn.Module):
    """YOLOv8 mask Proto module for segmentation models."""

    def __init__(self, c1, c_=256, c2=32):
        """
        Initializes the YOLOv8 mask Proto module with specified number of protos and masks.

        Input arguments are ch_in, number of protos, number of masks.
        """
        super().__init__()
        self.cv1 = Conv(c1, c_, k=3)
        self.upsample = nn.ConvTranspose2d(c_, c_, 2, 2, 0, bias=True)  # nn.Upsample(scale_factor=2, mode='nearest')
        self.cv2 = Conv(c_, c_, k=3)
        self.cv3 = Conv(c_, c2)

    def forward(self, x):
        """Performs a forward pass through layers using an upsampled input image."""
        return self.cv3(self.cv2(self.upsample(self.cv1(x))))


class HGStem(nn.Module):
    """
    StemBlock of PPHGNetV2 with 5 convolutions and one maxpool2d.

    https://github.com/PaddlePaddle/PaddleDetection/blob/develop/ppdet/modeling/backbones/hgnet_v2.py
    """

    def __init__(self, c1, cm, c2):
        """Initialize the SPP layer with input/output channels and specified kernel sizes for max pooling."""
        super().__init__()
        self.stem1 = Conv(c1, cm, 3, 2, act=nn.ReLU())
        self.stem2a = Conv(cm, cm // 2, 2, 1, 0, act=nn.ReLU())
        self.stem2b = Conv(cm // 2, cm, 2, 1, 0, act=nn.ReLU())
        self.stem3 = Conv(cm * 2, cm, 3, 2, act=nn.ReLU())
        self.stem4 = Conv(cm, c2, 1, 1, act=nn.ReLU())
        self.pool = nn.MaxPool2d(kernel_size=2, stride=1, padding=0, ceil_mode=True)

    def forward(self, x):
        """Forward pass of a PPHGNetV2 backbone layer."""
        x = self.stem1(x)
        x = F.pad(x, [0, 1, 0, 1])
        x2 = self.stem2a(x)
        x2 = F.pad(x2, [0, 1, 0, 1])
        x2 = self.stem2b(x2)
        x1 = self.pool(x)
        x = torch.cat([x1, x2], dim=1)
        x = self.stem3(x)
        x = self.stem4(x)
        return x


class HGBlock(nn.Module):
    """
    HG_Block of PPHGNetV2 with 2 convolutions and LightConv.

    https://github.com/PaddlePaddle/PaddleDetection/blob/develop/ppdet/modeling/backbones/hgnet_v2.py
    """

    def __init__(self, c1, cm, c2, k=3, n=6, lightconv=False, shortcut=False, act=nn.ReLU()):
        """Initializes a CSP Bottleneck with 1 convolution using specified input and output channels."""
        super().__init__()
        block = LightConv if lightconv else Conv
        self.m = nn.ModuleList(block(c1 if i == 0 else cm, cm, k=k, act=act) for i in range(n))
        self.sc = Conv(c1 + n * cm, c2 // 2, 1, 1, act=act)  # squeeze conv
        self.ec = Conv(c2 // 2, c2, 1, 1, act=act)  # excitation conv
        self.add = shortcut and c1 == c2

    def forward(self, x):
        """Forward pass of a PPHGNetV2 backbone layer."""
        y = [x]
        y.extend(m(y[-1]) for m in self.m)
        y = self.ec(self.sc(torch.cat(y, 1)))
        return y + x if self.add else y


class SPP(nn.Module):
    """Spatial Pyramid Pooling (SPP) layer https://arxiv.org/abs/1406.4729."""

    def __init__(self, c1, c2, k=(5, 9, 13)):
        """Initialize the SPP layer with input/output channels and pooling kernel sizes."""
        super().__init__()
        c_ = c1 // 2  # hidden channels
        self.cv1 = Conv(c1, c_, 1, 1)
        self.cv2 = Conv(c_ * (len(k) + 1), c2, 1, 1)
        self.m = nn.ModuleList([nn.MaxPool2d(kernel_size=x, stride=1, padding=x // 2) for x in k])

    def forward(self, x):
        """Forward pass of the SPP layer, performing spatial pyramid pooling."""
        x = self.cv1(x)
        return self.cv2(torch.cat([x] + [m(x) for m in self.m], 1))


class SPPF(nn.Module):
    """Spatial Pyramid Pooling - Fast (SPPF) layer for YOLOv5 by Glenn Jocher."""

    def __init__(self, c1, c2, k=5):
        """
        Initializes the SPPF layer with given input/output channels and kernel size.

        This module is equivalent to SPP(k=(5, 9, 13)).
        """
        super().__init__()
        c_ = c1 // 2  # hidden channels
        self.cv1 = Conv(c1, c_, 1, 1)
        self.cv2 = Conv(c_ * 4, c2, 1, 1)
        self.m = nn.MaxPool2d(kernel_size=k, stride=1, padding=k // 2)

    def forward(self, x):
        """Forward pass through Ghost Convolution block."""
        x = self.cv1(x)
        y1 = self.m(x)
        y2 = self.m(y1)
        return self.cv2(torch.cat((x, y1, y2, self.m(y2)), 1))


class C1(nn.Module):
    """CSP Bottleneck with 1 convolution."""

    def __init__(self, c1, c2, n=1):
        """Initializes the CSP Bottleneck with configurations for 1 convolution with arguments ch_in, ch_out, number."""
        super().__init__()
        self.cv1 = Conv(c1, c2, 1, 1)
        self.m = nn.Sequential(*(Conv(c2, c2, 3) for _ in range(n)))

    def forward(self, x):
        """Applies cross-convolutions to input in the C3 module."""
        y = self.cv1(x)
        return self.m(y) + y


class C2(nn.Module):
    """CSP Bottleneck with 2 convolutions."""

    def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):
        """Initializes the CSP Bottleneck with 2 convolutions module with arguments ch_in, ch_out, number, shortcut,
        groups, expansion.
        """
        super().__init__()
        self.c = int(c2 * e)  # hidden channels
        self.cv1 = Conv(c1, 2 * self.c, 1, 1)
        self.cv2 = Conv(2 * self.c, c2, 1)  # optional act=FReLU(c2)
        # self.attention = ChannelAttention(2 * self.c)  # or SpatialAttention()
        self.m = nn.Sequential(*(Bottleneck(self.c, self.c, shortcut, g, k=((3, 3), (3, 3)), e=1.0) for _ in range(n)))

    def forward(self, x):
        """Forward pass through the CSP bottleneck with 2 convolutions."""
        a, b = self.cv1(x).chunk(2, 1)
        return self.cv2(torch.cat((self.m(a), b), 1))


class C2f(nn.Module):
    """Faster Implementation of CSP Bottleneck with 2 convolutions."""

    def __init__(self, c1, c2, n=1, shortcut=False, g=1, e=0.5):
        """Initialize CSP bottleneck layer with two convolutions with arguments ch_in, ch_out, number, shortcut, groups,
        expansion.
        """
        super().__init__()
        self.c = int(c2 * e)  # hidden channels
        self.cv1 = Conv(c1, 2 * self.c, 1, 1)
        self.cv2 = Conv((2 + n) * self.c, c2, 1)  # optional act=FReLU(c2)
        self.m = nn.ModuleList(Bottleneck(self.c, self.c, shortcut, g, k=((3, 3), (3, 3)), e=1.0) for _ in range(n))

    def forward(self, x):
        """Forward pass through C2f layer."""
        y = list(self.cv1(x).chunk(2, 1))
        y.extend(m(y[-1]) for m in self.m)
        return self.cv2(torch.cat(y, 1))

    def forward_split(self, x):
        """Forward pass using split() instead of chunk()."""
        y = list(self.cv1(x).split((self.c, self.c), 1))
        y.extend(m(y[-1]) for m in self.m)
        return self.cv2(torch.cat(y, 1))


class C3(nn.Module):
    """CSP Bottleneck with 3 convolutions."""

    def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):
        """Initialize the CSP Bottleneck with given channels, number, shortcut, groups, and expansion values."""
        super().__init__()
        c_ = int(c2 * e)  # hidden channels
        self.cv1 = Conv(c1, c_, 1, 1)
        self.cv2 = Conv(c1, c_, 1, 1)
        self.cv3 = Conv(2 * c_, c2, 1)  # optional act=FReLU(c2)
        self.m = nn.Sequential(*(Bottleneck(c_, c_, shortcut, g, k=((1, 1), (3, 3)), e=1.0) for _ in range(n)))

    def forward(self, x):
        """Forward pass through the CSP bottleneck with 2 convolutions."""
        return self.cv3(torch.cat((self.m(self.cv1(x)), self.cv2(x)), 1))


class C3x(C3):
    """C3 module with cross-convolutions."""

    def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):
        """Initialize C3TR instance and set default parameters."""
        super().__init__(c1, c2, n, shortcut, g, e)
        self.c_ = int(c2 * e)
        self.m = nn.Sequential(*(Bottleneck(self.c_, self.c_, shortcut, g, k=((1, 3), (3, 1)), e=1) for _ in range(n)))


class RepC3(nn.Module):
    """Rep C3."""

    def __init__(self, c1, c2, n=3, e=1.0):
        """Initialize CSP Bottleneck with a single convolution using input channels, output channels, and number."""
        super().__init__()
        c_ = int(c2 * e)  # hidden channels
        self.cv1 = Conv(c1, c2, 1, 1)
        self.cv2 = Conv(c1, c2, 1, 1)
        self.m = nn.Sequential(*[RepConv(c_, c_) for _ in range(n)])
        self.cv3 = Conv(c_, c2, 1, 1) if c_ != c2 else nn.Identity()

    def forward(self, x):
        """Forward pass of RT-DETR neck layer."""
        return self.cv3(self.m(self.cv1(x)) + self.cv2(x))


class C3TR(C3):
    """C3 module with TransformerBlock()."""

    def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):
        """Initialize C3Ghost module with GhostBottleneck()."""
        super().__init__(c1, c2, n, shortcut, g, e)
        c_ = int(c2 * e)
        self.m = TransformerBlock(c_, c_, 4, n)


class C3Ghost(C3):
    """C3 module with GhostBottleneck()."""

    def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):
        """Initialize 'SPP' module with various pooling sizes for spatial pyramid pooling."""
        super().__init__(c1, c2, n, shortcut, g, e)
        c_ = int(c2 * e)  # hidden channels
        self.m = nn.Sequential(*(GhostBottleneck(c_, c_) for _ in range(n)))


class GhostBottleneck(nn.Module):
    """Ghost Bottleneck https://github.com/huawei-noah/ghostnet."""

    def __init__(self, c1, c2, k=3, s=1):
        """Initializes GhostBottleneck module with arguments ch_in, ch_out, kernel, stride."""
        super().__init__()
        c_ = c2 // 2
        self.conv = nn.Sequential(
            GhostConv(c1, c_, 1, 1),  # pw
            DWConv(c_, c_, k, s, act=False) if s == 2 else nn.Identity(),  # dw
            GhostConv(c_, c2, 1, 1, act=False))  # pw-linear
        self.shortcut = nn.Sequential(DWConv(c1, c1, k, s, act=False), Conv(c1, c2, 1, 1,
                                                                            act=False)) if s == 2 else nn.Identity()

    def forward(self, x):
        """Applies skip connection and concatenation to input tensor."""
        return self.conv(x) + self.shortcut(x)


class Bottleneck(nn.Module):
    """Standard bottleneck."""

    def __init__(self, c1, c2, shortcut=True, g=1, k=(3, 3), e=0.5):
        """Initializes a bottleneck module with given input/output channels, shortcut option, group, kernels, and
        expansion.
        """
        super().__init__()
        c_ = int(c2 * e)  # hidden channels
        self.cv1 = Conv(c1, c_, k[0], 1)
        self.cv2 = Conv(c_, c2, k[1], 1, g=g)
        self.add = shortcut and c1 == c2

    def forward(self, x):
        """'forward()' applies the YOLO FPN to input data."""
        return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x))


class BottleneckCSP(nn.Module):
    """CSP Bottleneck https://github.com/WongKinYiu/CrossStagePartialNetworks."""

    def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):
        """Initializes the CSP Bottleneck given arguments for ch_in, ch_out, number, shortcut, groups, expansion."""
        super().__init__()
        c_ = int(c2 * e)  # hidden channels
        self.cv1 = Conv(c1, c_, 1, 1)
        self.cv2 = nn.Conv2d(c1, c_, 1, 1, bias=False)
        self.cv3 = nn.Conv2d(c_, c_, 1, 1, bias=False)
        self.cv4 = Conv(2 * c_, c2, 1, 1)
        self.bn = nn.BatchNorm2d(2 * c_)  # applied to cat(cv2, cv3)
        self.act = nn.SiLU()
        self.m = nn.Sequential(*(Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n)))

    def forward(self, x):
        """Applies a CSP bottleneck with 3 convolutions."""
        y1 = self.cv3(self.m(self.cv1(x)))
        y2 = self.cv2(x)
        return self.cv4(self.act(self.bn(torch.cat((y1, y2), 1))))



def conv_bn(in_channels, out_channels, kernel_size, stride, padding, groups=1):
    result = nn.Sequential()
    result.add_module('conv', nn.Conv2d(in_channels=in_channels, out_channels=out_channels,
                                        kernel_size=kernel_size, stride=stride, padding=padding, groups=groups,
                                        bias=False))
    result.add_module('bn', nn.BatchNorm2d(num_features=out_channels))

    return result

class SEBlock(nn.Module):

    def __init__(self, input_channels, internal_neurons):
        super(SEBlock, self).__init__()
        self.down = nn.Conv2d(in_channels=input_channels, out_channels=internal_neurons, kernel_size=1, stride=1,
                              bias=True)
        self.up = nn.Conv2d(in_channels=internal_neurons, out_channels=input_channels, kernel_size=1, stride=1,
                            bias=True)
        self.input_channels = input_channels

    def forward(self, inputs):
        x = F.avg_pool2d(inputs, kernel_size=inputs.size(3))
        x = self.down(x)
        x = F.relu(x)
        x = self.up(x)
        x = torch.sigmoid(x)
        x = x.view(-1, self.input_channels, 1, 1)
        return inputs * x

class RepVGGBlock(nn.Module):
    def __init__(self, in_channels, out_channels, kernel_size=3,
                 stride=1, padding=1, dilation=1, groups=1, padding_mode='zeros', deploy=False, use_se=False):
        super(RepVGGBlock, self).__init__()
        self.deploy = deploy
        self.groups = groups
        self.in_channels = in_channels
        padding_11 = padding - kernel_size // 2
        self.nonlinearity = nn.SiLU()
        # self.nonlinearity = nn.ReLU()
        if use_se:
            self.se = SEBlock(out_channels, internal_neurons=out_channels // 16)
        else:
            self.se = nn.Identity()
        if deploy:
            self.rbr_reparam = nn.Conv2d(in_channels=in_channels, out_channels=out_channels, kernel_size=kernel_size,
                                         stride=stride,
                                         padding=padding, dilation=dilation, groups=groups, bias=True,
                                         padding_mode=padding_mode)
 
        else:
            self.rbr_identity = nn.BatchNorm2d(
                num_features=in_channels) if out_channels == in_channels and stride == 1 else None
            self.rbr_dense = conv_bn(in_channels=in_channels, out_channels=out_channels, kernel_size=kernel_size,
                                     stride=stride, padding=padding, groups=groups)
            self.rbr_1x1 = conv_bn(in_channels=in_channels, out_channels=out_channels, kernel_size=1, stride=stride,
                                   padding=padding_11, groups=groups)
            # print('RepVGG Block, identity = ', self.rbr_identity)
    def switch_to_deploy(self):
        if hasattr(self, 'rbr_1x1'):
            kernel, bias = self.get_equivalent_kernel_bias()
            self.rbr_reparam = nn.Conv2d(in_channels=self.rbr_dense.conv.in_channels, out_channels=self.rbr_dense.conv.out_channels,
                                    kernel_size=self.rbr_dense.conv.kernel_size, stride=self.rbr_dense.conv.stride,
                                    padding=self.rbr_dense.conv.padding, dilation=self.rbr_dense.conv.dilation, groups=self.rbr_dense.conv.groups, bias=True)
            self.rbr_reparam.weight.data = kernel
            self.rbr_reparam.bias.data = bias
            for para in self.parameters():
                para.detach_()
            self.rbr_dense = self.rbr_reparam
            # self.__delattr__('rbr_dense')
            self.__delattr__('rbr_1x1')
            if hasattr(self, 'rbr_identity'):
                self.__delattr__('rbr_identity')
            if hasattr(self, 'id_tensor'):
                self.__delattr__('id_tensor')
            self.deploy = True
 
    def get_equivalent_kernel_bias(self):
        kernel3x3, bias3x3 = self._fuse_bn_tensor(self.rbr_dense)
        kernel1x1, bias1x1 = self._fuse_bn_tensor(self.rbr_1x1)
        kernelid, biasid = self._fuse_bn_tensor(self.rbr_identity)
        return kernel3x3 + self._pad_1x1_to_3x3_tensor(kernel1x1) + kernelid, bias3x3 + bias1x1 + biasid
 
    def _pad_1x1_to_3x3_tensor(self, kernel1x1):
        if kernel1x1 is None:
            return 0
        else:
            return torch.nn.functional.pad(kernel1x1, [1, 1, 1, 1])
 
    def _fuse_bn_tensor(self, branch):
        if branch is None:
            return 0, 0
        if isinstance(branch, nn.Sequential):
            kernel = branch.conv.weight
            running_mean = branch.bn.running_mean
            running_var = branch.bn.running_var
            gamma = branch.bn.weight
            beta = branch.bn.bias
            eps = branch.bn.eps
        else:
            assert isinstance(branch, nn.BatchNorm2d)
            if not hasattr(self, 'id_tensor'):
                input_dim = self.in_channels // self.groups
                kernel_value = np.zeros((self.in_channels, input_dim, 3, 3), dtype=np.float32)
                for i in range(self.in_channels):
                    kernel_value[i, i % input_dim, 1, 1] = 1
                self.id_tensor = torch.from_numpy(kernel_value).to(branch.weight.device)
            kernel = self.id_tensor
            running_mean = branch.running_mean
            running_var = branch.running_var
            gamma = branch.weight
            beta = branch.bias
            eps = branch.eps
        std = (running_var + eps).sqrt()
        t = (gamma / std).reshape(-1, 1, 1, 1)
        return kernel * t, beta - running_mean * gamma / std
 
    def forward(self, inputs):
        if self.deploy:
            return self.nonlinearity(self.rbr_dense(inputs))
        if hasattr(self, 'rbr_reparam'):
            return self.nonlinearity(self.se(self.rbr_reparam(inputs)))
 
        if self.rbr_identity is None:
            id_out = 0
        else:
            id_out = self.rbr_identity(inputs)
        return self.nonlinearity(self.se(self.rbr_dense(inputs) + self.rbr_1x1(inputs) + id_out))

步骤 3：RepVGGBlock 模块注册
（路径位于 ultralytics/ultralytics/nn/modules/block.py）

ultralytics/ultralytics/nn/modules/init.py

步骤 4：修改 tasks.py 中的 parse_model 函数
（路径位于 ultralytics/ultralytics/nn/tasks.py）


步骤 5：python setup.py

3.3 三明治结构

注意：YOLOv8s 三明治 DW 通道为 L 的一半，同时论文 DW 卷积核大小存在疑点（无法进行特征图 concat 操作），均改用 3*3。
主要修改配置文件中 head 部分（图里通道 s 版本为 1/2）：

# Ultralytics YOLO 🚀, AGPL-3.0 license
# YOLOv8 object detection model with P2-P5 outputs. For Usage examples see https://docs.ultralytics.com/tasks/detect

# Parameters
nc: 80  # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n.yaml' will call yolov8.yaml with scale 'n'
  # [depth, width, max_channels]
  n: [0.33, 0.25, 1024]
  s: [0.33, 0.50, 1024]
  m: [0.67, 0.75, 768]
  l: [1.00, 1.00, 512]
  x: [1.00, 1.25, 512]

# YOLOv8.0 backbone
backbone:
  # [from, repeats, module, args]
  - [-1, 1, Conv, [64, 3, 2]]  # 0-P1/2
  - [-1, 1, RepVGGBlock, [128, 3, 2]]  # 1-P2/4
  - [-1, 3, C2f, [128, True]]
  - [-1, 1, RepVGGBlock, [256, 3, 2]]  # 3-P3/8
  - [-1, 6, C2f, [256, True]]
  - [-1, 1, RepVGGBlock, [512, 3, 2]]  # 5-P4/16
  - [-1, 6, C2f, [512, True]]
  - [-1, 1, RepVGGBlock, [1024, 3, 2]]  # 7-P5/32
  - [-1, 3, C2f, [1024, True]]
  - [-1, 1, SPPF, [1024, 5]]  # 9

# YOLOv8.0-p2 head
head:
  
  - [-1, 1, nn.Upsample, [None, 2, 'nearest']]  #10
  - [4, 1, DWConv, [128, 3, 2]] ################################## p3深度可分离卷积层后(v8s是c=64/k=3/s=2/p=1)
  - [[-1, 6, 10], 1, Concat, [1]]  # cat backbone P4
  - [-1, 3, C2f, [512]]  # 13

  
  - [-1, 1, nn.Upsample, [None, 2, 'nearest']]
  - [2, 1, DWConv, [64, 3, 2]] ################################## p2深度可分离卷积层加入融合(v8s是c=32/k=3/s=2/p=1)
  - [[-1, 4, 14], 1, Concat, [1]]  # cat backbone P3
  - [-1, 3, C2f, [256]]  # 17 (P3/8-small)

  
  - [-1, 1, nn.Upsample, [None, 2, 'nearest']]
  - [0, 1, DWConv, [32, 3, 2]] ################################## p1深度可分离卷积层加入融合(v8s是c=16/k=3/s=2/p=1)
  - [[-1, 2, 18], 1, Concat, [1]]  # cat backbone P2
  - [-1, 3, C2f, [128]]  # 21 (P2/4-xsmall)

  - [-1, 1, Conv, [128, 3, 2]]
  - [[-1, 17], 1, Concat, [1]]  # cat head P3
  - [-1, 3, C2f, [256]]  # 24 (P3/8-small)

  - [-1, 1, Conv, [256, 3, 2]]
  - [[-1, 13], 1, Concat, [1]]  # cat head P4
  - [-1, 3, C2f, [512]]  # 27 (P4/16-medium)

  - [-1, 1, Conv, [512, 3, 2]]
  - [[-1, 9], 1, Concat, [1]]  # cat head P5
  - [-1, 3, C2f, [1024]]  # 30 (P5/32-large)

  - [[21, 24, 27, 30], 1, Detect, [nc]]  # Detect(P2, P3, P4, P5)

4 YOLOv8s 版复现结果

注意：论文未用预训练权重！
新建train.py，内容如下：

import os
os.environ['CUDA_VISIBLE_DEVICES'] = '0'
from ultralytics import YOLO

# Load a model
# model = YOLO(model="yolov8s.yaml")  # yolov8s
# model = YOLO(model="yolov8s-p2.yaml")  # yolov8s+head
# model = YOLO(model="yolov8s-p2-repvgg.yaml")  # yolov8s+repvgg
model = YOLO(model="yolov8s-p2-repvgg-sf.yaml")  # yolov8s+repvgg+sf

# model.load('yolov8s.pt')  # 论文未加载预训练权重
# Use the model
model.train(data="VisDrone.yaml", imgsz=640, epochs=300, workers=8, batch=8, cache=True, project='runs/train')
# path = model.export(format="onnx", dynamic=True)  # export the mode l to ONNX format

4.1 复现结果：

针对无人机数据集 Visdrone 训练复现结果：接近。

5 结论

（1）性能主要提升来自增加对小目标检测的 P2 层（即增加一个 160*160 的大尺度检测头），因为无人机数据集小目标偏多，该 P2 层改进对小目标能够有效提升；
（2）增加 P2 分支与多支路融合虽然在模型参数方面略微增加，但推理耗时显著增加，不利于实时部署。
（3）加 P2 分支会对大目标识别精度略微降低【小目标较多的场景可以借鉴一下】；
（4）三明治改进几乎无效果，且增加推理耗时【不建议使用】；
（5）RepVGG在不带来推理耗时成本前提下，一般能够提升模型性能。