YOLOv11 改进 - 主干网络 EfficientRep：一种旨在提高硬件效率的RepVGG风格卷积神经网络架构

前言本文介绍了EfficientRep卷积神经网络架构及其核心模块在YOLOv11中的结合应用。为解决传统神经网络效率评估指标对硬件不敏感问题提出设计硬件感知神经网络的方法构建EfficientRep架构。该架构采用RepVGG风格卷积结构训练时结合多分支确保准确性推理时转换为单分支。我们将EfficientRep的RepVGGBlock和RepBlock集成进YOLOv11实验表明其在单GPU设备上提升了训练和推理速度改进后的YOLOv11也取得了良好实验结果。文章目录 YOLOv11改进大全卷积层、轻量化、注意力机制、损失函数、Backbone、SPPF、Neck、检测头全方位优化汇总专栏链接: YOLOv11改进专栏介绍摘要我们提出了一种硬件高效的卷积神经网络架构该架构采用类似RepVGG的结构设计范式。传统的网络效率评估指标如FLOPs或参数量对硬件特性包括计算能力与内存带宽缺乏敏感性因此如何设计能够有效利用硬件计算资源和内存带宽的神经网络架构成为关键科学问题。本文提出了一种硬件感知的神经网络设计方法基于该方法我们开发了EfficientRep系列卷积网络这些网络针对高性能计算硬件如图形处理器GPU进行了深度优化并成功应用于YOLOv6目标检测框架。YOLOv6框架已发布v1和v2两个版本包含YOLOv6N、YOLOv6S、YOLOv6M和YOLOv6L等多个不同规模的模型变体。相关实现代码已开源可通过https://github.com/meituan/YOLOv6获取。文章链接论文地址论文地址代码地址代码地址基本原理EfficientRep是一种高效的RepVGG风格卷积神经网络架构旨在优化硬件的计算能力和内存带宽利用。该架构采用RepVGG风格的卷积结构具有3x3卷积核通过Winograd算法在GPU或CPU上进行高度优化。在训练状态下EfficientRep结合了3x3分支、1x1分支和恒等映射以确保训练期间的准确性。在推理状态下通过重新参数化多分支结构转换为单分支的3x3卷积。EfficientRep主要包括EfficientRep骨干网络和Rep-PAN颈部这些结构对GPU友好并应用于YOLOv6检测框架YOLOv6-v1[T3]。在YOLOv6-v1中EfficientRep骨干网络和Rep-PAN颈部的设计使得单GPU设备上的训练和推理速度得到提升。然而当YOLOv6-v1扩展到中等规模时推理速度下降过快准确性也无法与CSP风格的YOLO系列相竞争。因此为了在大型模型中实现更好的准确性和速度平衡研究人员探索了多路径结构等新颖设计[T6]。EfficientRep将在backbone中stride2的卷积层换成了stride2的RepConv层。并且也将CSP-Block修改为了RepBlock。核心代码import torch import torch.nn as nn # 定义卷积和批归一化的组合函数 def conv_bn(in_channels, out_channels, kernel_size, stride, padding, groups1): result nn.Sequential() # 使用 nn.Sequential 将多个层组合 # 添加卷积层 result.add_module(conv, nn.Conv2d(in_channelsin_channels, out_channelsout_channels, kernel_sizekernel_size, stridestride, paddingpadding, groupsgroups, biasFalse)) # 添加批归一化层 result.add_module(bn, nn.BatchNorm2d(num_featuresout_channels)) return result # 定义 RepVGGBlock 类继承自 nn.Module class RepVGGBlock(nn.Module): def __init__(self, in_channels, out_channels, kernel_size3, stride1, padding1, dilation1, groups1, padding_modezeros, deployFalse, use_seFalse): super(RepVGGBlock, self).__init__() self.deploy deploy self.groups groups self.in_channels in_channels padding_11 padding - kernel_size // 2 # 使用 SiLU 作为激活函数 self.nonlinearity nn.SiLU() self.se nn.Identity() # 使用 Identity 模块 if deploy: # 如果处于部署模式使用单一的卷积层 self.rbr_reparam nn.Conv2d(in_channelsin_channels, out_channelsout_channels, kernel_sizekernel_size, stridestride, paddingpadding, dilationdilation, groupsgroups, biasTrue, padding_modepadding_mode) else: # 否则使用多分支结构 self.rbr_identity nn.BatchNorm2d(num_featuresin_channels) if out_channels in_channels and stride 1 else None self.rbr_dense conv_bn(in_channelsin_channels, out_channelsout_channels, kernel_sizekernel_size, stridestride, paddingpadding, groupsgroups) self.rbr_1x1 conv_bn(in_channelsin_channels, out_channelsout_channels, kernel_size1, stridestride, paddingpadding_11, groupsgroups) # 融合批归一化层与卷积层的权重 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), dtypenp.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 # 获取等效的卷积核和偏置 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 # 将1x1卷积核填充到3x3大小 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]) # 前向传播函数 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)) # 定义 RepBlock 类继承自 nn.Module class RepBlock(nn.Module): RepBlock 是一个包含多个 RepVGGBlock 的阶段块 def __init__(self, in_channels, out_channels, n1, isTrueNone): super().__init__() self.conv1 RepVGGBlock(in_channels, out_channels) self.block nn.Sequential(*(RepVGGBlock(out_channels, out_channels) for _ in range(n - 1))) if n 1 else None # 前向传播函数 def forward(self, x): x self.conv1(x) if self.block is not None: x self.block(x) return x实验脚本import warnings warnings.filterwarnings(ignore) from ultralytics import YOLO if __name__ __main__: # 修改为自己的配置文件地址 model YOLO(/root/ultralytics-main/ultralytics/cfg/models/11/yolov11-EfficientRep.yaml) # 修改为自己的数据集地址 model.train(data/root/ultralytics-main/ultralytics/cfg/datasets/coco8.yaml, cacheFalse, imgsz640, epochs10, single_clsFalse, # 是否是单类别检测 batch8, close_mosaic10, workers0, optimizerSGD, ampTrue, projectruns/train, nameEfficientRep, )结果