【无标题】YOLOV5和YOLOV8主干改进：华为诺亚提出全新骨干架构VanillaNet

VanillaNet：一种极简设计的神经网络架构，性能与代表性CNN工作和视觉Transformer不相上下，突出了极简主义在深度学习中的潜力。

单位：华为诺亚, 悉尼大学

代码：https://github.com/huawei-noah/VanillaNet

论文：https://arxiv.org/abs/2305.12972

在本研究中，我们介绍了 VanillaNet，这是一种设计优雅的神经网络架构。 通过避免高深度、shortcuts和自注意力等复杂操作，VanillaNet 简洁明了但功能强大。

每一层都经过精心设计，简洁明了，非线性激活函数在训练后被剪枝以恢复原始架构。VanillaNet 克服了固有复杂性的挑战，使其成为资源受限环境的理想选择。 其易于理解和高度简化的架构为高效部署开辟了新的可能性。

在这里，我们以6层为例详细展示了VanillaNet的体系结构。对于主干，我们使用4×4×3×C卷积层，步长为4，遵循[18，31，32]中的流行设置，将具有3个通道的图像映射到具有C个通道的特征。在阶段1、2和3，使用步长为2的最大池化层来减小大小和特征图，并且通道的数量增加2。在第4阶段，我们不增加信道的数量，因为它遵循平均池化层。最后一层是完全连接的层，用于输出分类结果。每个卷积层的核大小是1×1，因为我们的目标是在保留特征图信息的同时，对每个层使用最小的计算成本。在每个1×1卷积层之后应用激活函数。为了简化网络的训练过程，还在每一层之后添加了批量规范化。对于不同层数的VanillaNet，我们在每个阶段添加块，这将在补充材料中详细说明。需要注意的是，VanillaNet没有捷径，因为我们从经验上发现，添加捷径几乎不会提高性能。这也带来了另一个好处，即所提出的架构非常容易实现，因为没有分支和额外的块，如挤压和激励块[22]。

 实验结果

广泛的实验表明，VanillaNet 提供的性能与著名的深度神经网络和视觉Transformer器相当，展示了深度学习中极简主义的力量

github 官网下载vanillanet.py

#Copyright (C) 2023. Huawei Technologies Co., Ltd. All rights reserved.

#This program is free software; you can redistribute it and/or modify it under the terms of the MIT License.

#This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the MIT License for more details.

import torch
import torch.nn as nn
import torch.nn.functional as F
from timm.models.layers import weight_init, DropPath
from timm.models.registry import register_model


class activation(nn.ReLU):
    # 初始化激活函数
    def __init__(self, dim, act_num=3, deploy=False):
        super(activation, self).__init__()
        # 判断是否是部署模式
        self.deploy = deploy
        # 初始化权重
        self.weight = torch.nn.Parameter(torch.randn(dim, 1, act_num*2 + 1, act_num*2 + 1))
        # 初始化偏置
        self.bias = None
        # 初始化BatchNorm2d
        self.bn = nn.BatchNorm2d(dim, eps=1e-6)
        # 设置维度
        self.dim = dim
        # 设置激活函数个数
        self.act_num = act_num
        # 初始化权重
        weight_init.trunc_normal_(self.weight, std=.02)

    def forward(self, x):
        if self.deploy:
            return torch.nn.functional.conv2d(
                super(activation, self).forward(x), 
                self.weight, self.bias, padding=(self.act_num*2 + 1)//2, groups=self.dim)
        else:
            return self.bn(torch.nn.functional.conv2d(
                super(activation, self).forward(x),
                self.weight, padding=(self.act_num*2 + 1)//2, groups=self.dim))

    def _fuse_bn_tensor(self, weight, bn):
        kernel = weight
        running_mean = bn.running_mean
        running_var = bn.running_var
        gamma = bn.weight
        beta = bn.bias
        eps = bn.eps
        std = (running_var + eps).sqrt()
        t = (gamma / std).reshape(-1, 1, 1, 1)
        return kernel * t, beta + (0 - running_mean) * gamma / std
    
    def switch_to_deploy(self):
        kernel, bias = self._fuse_bn_tensor(self.weight, self.bn)
        self.weight.data = kernel
        self.bias = torch.nn.Parameter(torch.zeros(self.dim))
        self.bias.data = bias
        self.__delattr__('bn')
        self.deploy = True


class vanillanetBlock(nn.Module):
    # 初始化vanillanetBlock类
    def __init__(self, dim, dim_out, act_num=3, stride=2, deploy=False, ada_pool=None):
        super().__init__()
        # 定义激活函数
        self.act_learn = 1
        self.deploy = deploy
        # 如果deploy为True，则将conv设置为Conv2d，否则设置为Conv2d + BatchNorm2d
        if self.deploy:
            self.conv = nn.Conv2d(dim, dim_out, kernel_size=1)
        else:
            self.conv1 = nn.Sequential(
                nn.Conv2d(dim, dim, kernel_size=1),
                nn.BatchNorm2d(dim, eps=1e-6),
            )
            self.conv2 = nn.Sequential(
                nn.Conv2d(dim, dim_out, kernel_size=1),
                nn.BatchNorm2d(dim_out, eps=1e-6)
            )

        # 如果不是ada_pool，则将pool设置为Identity，否则设置为AdaptiveMaxPool2d
        if not ada_pool:
            self.pool = nn.Identity() if stride == 1 else nn.MaxPool2d(stride)
        else:
            self.pool = nn.Identity() if stride == 1 else nn.AdaptiveMaxPool2d((ada_pool, ada_pool))

        # 定义激活函数
        self.act = activation(dim_out, act_num)
 
    def forward(self, x):
        if self.deploy:
            x = self.conv(x)
        else:
            x = self.conv1(x)
            x = torch.nn.functional.leaky_relu(x,self.act_learn)
            x = self.conv2(x)

        x = self.pool(x)
        x = self.act(x)
        return x

    def _fuse_bn_tensor(self, conv, bn):
        kernel = conv.weight
        bias = conv.bias
        running_mean = bn.running_mean
        running_var = bn.running_var
        gamma = bn.weight
        beta = bn.bias
        eps = bn.eps
        std = (running_var + eps).sqrt()
        t = (gamma / std).reshape(-1, 1, 1, 1)
        return kernel * t, beta + (bias - running_mean) * gamma / std
    
    def switch_to_deploy(self):
        kernel, bias = self._fuse_bn_tensor(self.conv1[0], self.conv1[1])
        self.conv1[0].weight.data = kernel
        self.conv1[0].bias.data = bias
        # kernel, bias = self.conv2[0].weight.data, self.conv2[0].bias.data
        kernel, bias = self._fuse_bn_tensor(self.conv2[0], self.conv2[1])
        self.conv = self.conv2[0]
        self.conv.weight.data = torch.matmul(kernel.transpose(1,3), self.conv1[0].weight.data.squeeze(3).squeeze(2)).transpose(1,3)
        self.conv.bias.data = bias + (self.conv1[0].bias.data.view(1,-1,1,1)*kernel).sum(3).sum(2).sum(1)
        self.__delattr__('conv1')
        self.__delattr__('conv2')
        self.act.switch_to_deploy()
        self.deploy = True
    

class VanillaNet(nn.Module):
    def __init__(self, in_chans=3, num_classes=1000, dims=[96, 192, 384, 768], 
                 drop_rate=0, act_num=3, strides=[2,2,2,1], deploy=False, ada_pool=None, **kwargs):
        super().__init__()
        self.deploy = deploy
        if self.deploy:
            self.stem = nn.Sequential(
                nn.Conv2d(in_chans, dims[0], kernel_size=4, stride=4),
                activation(dims[0], act_num)
            )
        else:
            self.stem1 = nn.Sequential(
                nn.Conv2d(in_chans, dims[0], kernel_size=4, stride=4),
                nn.BatchNorm2d(dims[0], eps=1e-6),
            )
            self.stem2 = nn.Sequential(
                nn.Conv2d(dims[0], dims[0], kernel_size=1, stride=1),
                nn.BatchNorm2d(dims[0], eps=1e-6),
                activation(dims[0], act_num)
            )

        self.act_learn = 1

        self.stages = nn.ModuleList()
        for i in range(len(strides)):
            if not ada_pool:
                stage = vanillanetBlock(dim=dims[i], dim_out=dims[i+1], act_num=act_num, stride=strides[i], deploy=deploy)
            else:
                stage = vanillanetBlock(dim=dims[i], dim_out=dims[i+1], act_num=act_num, stride=strides[i], deploy=deploy, ada_pool=ada_pool[i])
            self.stages.append(stage)
        self.depth = len(strides)

        if self.deploy:
            self.cls = nn.Sequential(
                nn.AdaptiveAvgPool2d((1,1)),
                nn.Dropout(drop_rate),
                nn.Conv2d(dims[-1], num_classes, 1),
            )
        else:
            self.cls1 = nn.Sequential(
                nn.AdaptiveAvgPool2d((1,1)),
                nn.Dropout(drop_rate),
                nn.Conv2d(dims[-1], num_classes, 1),
                nn.BatchNorm2d(num_classes, eps=1e-6),
            )
            self.cls2 = nn.Sequential(
                nn.Conv2d(num_classes, num_classes, 1)
            )
        
        self.apply(self._init_weights)

    def _init_weights(self, m):
        if isinstance(m, (nn.Conv2d, nn.Linear)):
            weight_init.trunc_normal_(m.weight, std=.02)
            nn.init.constant_(m.bias, 0)

    def change_act(self, m):
        for i in range(self.depth):
            self.stages[i].act_learn = m
        self.act_learn = m

    def forward(self, x):
        if self.deploy:
            x = self.stem(x)
        else:
            x = self.stem1(x)
            x = torch.nn.functional.leaky_relu(x,self.act_learn)
            x = self.stem2(x)

        for i in range(self.depth):
            x = self.stages[i](x)

        if self.deploy:
            x = self.cls(x)
        else:
            x = self.cls1(x)
            x = torch.nn.functional.leaky_relu(x,self.act_learn)
            x = self.cls2(x)
        return x.view(x.size(0),-1)

    def _fuse_bn_tensor(self, conv, bn):
        kernel = conv.weight
        bias = conv.bias
        running_mean = bn.running_mean
        running_var = bn.running_var
        gamma = bn.weight
        beta = bn.bias
        eps = bn.eps
        std = (running_var + eps).sqrt()
        t = (gamma / std).reshape(-1, 1, 1, 1)
        return kernel * t, beta + (bias - running_mean) * gamma / std
    
    def switch_to_deploy(self):
        self.stem2[2].switch_to_deploy()
        kernel, bias = self._fuse_bn_tensor(self.stem1[0], self.stem1[1])
        self.stem1[0].weight.data = kernel
        self.stem1[0].bias.data = bias
        kernel, bias = self._fuse_bn_tensor(self.stem2[0], self.stem2[1])
        self.stem1[0].weight.data = torch.einsum('oi,icjk->ocjk', kernel.squeeze(3).squeeze(2), self.stem1[0].weight.data)
        self.stem1[0].bias.data = bias + (self.stem1[0].bias.data.view(1,-1,1,1)*kernel).sum(3).sum(2).sum(1)
        self.stem = torch.nn.Sequential(*[self.stem1[0], self.stem2[2]])
        self.__delattr__('stem1')
        self.__delattr__('stem2')

        for i in range(self.depth):
            self.stages[i].switch_to_deploy()

        kernel, bias = self._fuse_bn_tensor(self.cls1[2], self.cls1[3])
        self.cls1[2].weight.data = kernel
        self.cls1[2].bias.data = bias
        kernel, bias = self.cls2[0].weight.data, self.cls2[0].bias.data
        self.cls1[2].weight.data = torch.matmul(kernel.transpose(1,3), self.cls1[2].weight.data.squeeze(3).squeeze(2)).transpose(1,3)
        self.cls1[2].bias.data = bias + (self.cls1[2].bias.data.view(1,-1,1,1)*kernel).sum(3).sum(2).sum(1)
        self.cls = torch.nn.Sequential(*self.cls1[0:3])
        self.__delattr__('cls1')
        self.__delattr__('cls2')
        self.deploy = True


@register_model
def vanillanet_5(pretrained=False,in_22k=False, **kwargs):
    model = VanillaNet(dims=[128*4, 256*4, 512*4, 1024*4], strides=[2,2,2], **kwargs)
    return model

@register_model
def vanillanet_6(pretrained=False,in_22k=False, **kwargs):
    model = VanillaNet(dims=[128*4, 256*4, 512*4, 1024*4, 1024*4], strides=[2,2,2,1], **kwargs)
    return model

@register_model
def vanillanet_7(pretrained=False,in_22k=False, **kwargs):
    model = VanillaNet(dims=[128*4, 128*4, 256*4, 512*4, 1024*4, 1024*4], strides=[1,2,2,2,1], **kwargs)
    return model

@register_model
def vanillanet_8(pretrained=False, in_22k=False, **kwargs):
    model = VanillaNet(dims=[128*4, 128*4, 256*4, 512*4, 512*4, 1024*4, 1024*4], strides=[1,2,2,1,2,1], **kwargs)
    return model

@register_model
def vanillanet_9(pretrained=False, in_22k=False, **kwargs):
    model = VanillaNet(dims=[128*4, 128*4, 256*4, 512*4, 512*4, 512*4, 1024*4, 1024*4], strides=[1,2,2,1,1,2,1], **kwargs)
    return model

@register_model
def vanillanet_10(pretrained=False, in_22k=False, **kwargs):
    model = VanillaNet(
        dims=[128*4, 128*4, 256*4, 512*4, 512*4, 512*4, 512*4, 1024*4, 1024*4],
        strides=[1,2,2,1,1,1,2,1],
        **kwargs)
    return model

@register_model
def vanillanet_11(pretrained=False, in_22k=False, **kwargs):
    model = VanillaNet(
        dims=[128*4, 128*4, 256*4, 512*4, 512*4, 512*4, 512*4, 512*4, 1024*4, 1024*4],
        strides=[1,2,2,1,1,1,1,2,1],
        **kwargs)
    return model

@register_model
def vanillanet_12(pretrained=False, in_22k=False, **kwargs):
    model = VanillaNet(
        dims=[128*4, 128*4, 256*4, 512*4, 512*4, 512*4, 512*4, 512*4, 512*4, 1024*4, 1024*4],
        strides=[1,2,2,1,1,1,1,1,2,1],
        **kwargs)
    return model

@register_model
def vanillanet_13(pretrained=False, in_22k=False, **kwargs):
    '''
    VanillaNet-13
    :param pretrained: 是否使用预训练模型
    :param in_22k: 是否使用22k帧的图像
    :param kwargs: 其他参数
    :return: VanillaNet
    '''
    model = VanillaNet(
        dims=[128*4, 128*4, 256*4, 512*4, 512*4, 512*4, 512*4, 512*4, 512*4, 512*4, 1024*4, 1024*4],
        strides=[1,2,2,1,1,1,1,1,1,2,1],
        **kwargs)
    return model

@register_model
def vanillanet_13_x1_5(pretrained=False, in_22k=False, **kwargs):
    model = VanillaNet(
        dims=[128*6, 128*6, 256*6, 512*6, 512*6, 512*6, 512*6, 512*6, 512*6, 512*6, 1024*6, 1024*6],
        strides=[1,2,2,1,1,1,1,1,1,2,1],
        **kwargs)
    return model

@register_model
def vanillanet_13_x1_5_ada_pool(pretrained=False, in_22k=False, **kwargs):
    model = VanillaNet(
        dims=[128*6, 128*6, 256*6, 512*6, 512*6, 512*6, 512*6, 512*6, 512*6, 512*6, 1024*6, 1024*6],
        strides=[1,2,2,1,1,1,1,1,1,2,1],
        ada_pool=[0,40,20,0,0,0,0,0,0,10,0],
        **kwargs)
    return model

在tasks.py中

from ultralytics.nn.vanillanet import (vanillanetBlock)

elif m is vanillanetBlock:
             c1, c2 = ch[f], args[0]
             if c2 != torch.NoneType:  
                 c2 = make_divisible(c2 * width, 8)
             args = [c1, c2, *args[1:]]

修改yaml文件

# Ultralytics YOLO 🚀, GPL-3.0 license

# Parameters
nc: 80  # number of classes
depth_multiple: 1  # scales module repeats
width_multiple: 0.5  # scales convolution channels

# YOLOv8.0n backbone
backbone:
  # [from, repeats, module, args]
  - [-1, 1, Conv, [64, 4, 4 ]] # 0-P2/4
  - [-1, 1, vanillanetBlock, [128, 1, 2]]  # 1-P2/8
  - [-1, 1, vanillanetBlock, [256, 1, 2]]  # 2-P2/16
  - [-1, 1, vanillanetBlock, [512, 1, 2]]  # 3-P2/32
  - [-1, 1, vanillanetBlock, [512, 1, 1]]  # 4-P3/32
  
# YOLOv8.0s head
head:
  - [-1, 1, nn.Upsample, [None, 2, 'nearest']] #1/16
  - [[-1, 2], 1, Concat, [1]]  # cat backbone P4
  - [-1, 3, C2f, [512]]  # 7

  - [-1, 1, nn.Upsample, [None, 2, 'nearest']] #1/8
  - [[-1, 1], 1, Concat, [1]]  # cat backbone P3
  - [-1, 3, C2f, [256]]  # 10 (P3/8-small)

  - [-1, 1, Conv, [256, 3, 2]] #1/16
  - [[-1, 7], 1, Concat, [1]]  # cat head P4
  - [-1, 3, C2f, [512]]  # 13 (P4/16-medium)

  - [-1, 1, Conv, [512, 3, 2]]
  - [[-1, 4], 1, Concat, [1]]  # cat head P5
  - [-1, 3, C2f, [1024]]  # 16 (P5/32-large)

  - [[10, 13, 16], 1, Detect, [nc]]  # Detect(P3, P4, P5)

运行训练即可