YOLO26改进| 特征融合 | 小波变换的多尺度特征融合

💡💡💡本专栏所有程序均经过测试，可成功执行💡💡💡

本文给大家带来的教程是将YOLO26的特征融合替换为WFU来提取特征。文章在介绍主要的原理后，将手把手教学如何进行模块的代码添加和修改，并将修改后的完整代码放在文章的最后，方便大家一键运行，小白也可轻松上手实践。以帮助您更好地学习深度学习目标检测YOLO系列的挑战。

专栏地址：YOLO26改进-论文涨点------点击跳转看所有内容，关注不迷路！****

1.论文

[2. WFU代码实现](#2. WFU代码实现)

[2.1 将WFU添加到YOLO26中](#2.1 将WFU添加到YOLO26中)

[2.2 更改init.py文件](#2.2 更改init.py文件)

[2.3 添加yaml文件](#2.3 添加yaml文件)

[2.4 在task.py中进行注册](#2.4 在task.py中进行注册)

[2.5 执行程序](#2.5 执行程序)

[3. 完整代码分享](#3. 完整代码分享)

[4. GFLOPs](#4. GFLOPs)

[5. 进阶](#5. 进阶)

6.总结

1.论文

论文地址： Efficient Face Super-Resolution via Wavelet-based Feature Enhancement Network

官方代码： 官方代码仓库点击即可跳转

2. WFU代码实现

2.1 将WFU添加到YOLO26中

**关键步骤一：**在ultralytics\ultralytics\nn\modules下面新建文件夹models，在文件夹下新建WFU.py，粘贴下面代码

python 复制代码

import torch
import torch.nn as nn
import torch.nn.functional as F
 
from ultralytics.nn.modules.conv import Conv     

class HaarWavelet(nn.Module):     
    def __init__(self, in_channels, grad=False):     
        super(HaarWavelet, self).__init__() 
        self.in_channels = in_channels

        self.haar_weights = torch.ones(4, 1, 2, 2)
        #h  
        self.haar_weights[1, 0, 0, 1] = -1
        self.haar_weights[1, 0, 1, 1] = -1 
        #v  
        self.haar_weights[2, 0, 1, 0] = -1
        self.haar_weights[2, 0, 1, 1] = -1   
        #d 
        self.haar_weights[3, 0, 1, 0] = -1
        self.haar_weights[3, 0, 0, 1] = -1 

        self.haar_weights = torch.cat([self.haar_weights] * self.in_channels, 0)    
        self.haar_weights = nn.Parameter(self.haar_weights)     
        self.haar_weights.requires_grad = grad

    def forward(self, x, rev=False):  
        if not rev:  
            out = F.conv2d(x, self.haar_weights, bias=None, stride=2, groups=self.in_channels) / 4.0
            out = out.reshape([x.shape[0], self.in_channels, 4, x.shape[2] // 2, x.shape[3] // 2]) 
            out = torch.transpose(out, 1, 2)     
            out = out.reshape([x.shape[0], self.in_channels * 4, x.shape[2] // 2, x.shape[3] // 2])
            return out  
        else:    
            out = x.reshape([x.shape[0], 4, self.in_channels, x.shape[2], x.shape[3]])
            out = torch.transpose(out, 1, 2)
            out = out.reshape([x.shape[0], self.in_channels * 4, x.shape[2], x.shape[3]])  
            return F.conv_transpose2d(out, self.haar_weights, bias=None, stride=2, groups = self.in_channels)
  
class WFU(nn.Module):
    def __init__(self, in_chn, ou_chn): 
        super(WFU, self).__init__()   
        dim_big, dim_small = in_chn    
        self.dim = dim_big    
        self.HaarWavelet = HaarWavelet(dim_big, grad=False) 
        self.InverseHaarWavelet = HaarWavelet(dim_big, grad=False) 
        self.RB = nn.Sequential(
            nn.Conv2d(dim_big, dim_big, kernel_size=3, padding=1),
            nn.ReLU(),     
            nn.Conv2d(dim_big, dim_big, kernel_size=3, padding=1),
        )
 
        self.channel_tranformation = nn.Sequential(     
            nn.Conv2d(dim_big+dim_small, dim_big+dim_small // 1, kernel_size=1, padding=0),    
            nn.ReLU(),
            nn.Conv2d(dim_big+dim_small // 1, dim_big*3, kernel_size=1, padding=0),
        )
     
        self.conv1x1 = Conv(dim_big, ou_chn) if dim_big != ou_chn else nn.Identity()
    
    def forward(self, x):
        x_big, x_small = x   
        haar = self.HaarWavelet(x_big, rev=False)
        a = haar.narrow(1, 0, self.dim) 
        h = haar.narrow(1, self.dim, self.dim)
        v = haar.narrow(1, self.dim*2, self.dim) 
        d = haar.narrow(1, self.dim*3, self.dim)     
     
        hvd = self.RB(h + v + d)
        a_ = self.channel_tranformation(torch.cat([x_small, a], dim=1)) 
        out = self.InverseHaarWavelet(torch.cat([hvd, a_], dim=1), rev=True)
        return self.conv1x1(out)

2.2 更改init.py文件

**关键步骤二：**在文件ultralytics\ultralytics\nn\modules\models文件夹下新建__init__.py文件，先导入函数

然后在下面的__all__中声明函数

2.3 添加yaml文件

**关键步骤三：**在/ultralytics/ultralytics/cfg/models/26下面新建文件yolo26_WFU.yaml文件，粘贴下面的内容

目标检测

python 复制代码

# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license

# Ultralytics YOLO26 object detection model with P3/8 - P5/32 outputs
# Model docs: https://docs.ultralytics.com/models/yolo26
# Task docs: https://docs.ultralytics.com/tasks/detect

# Parameters
nc: 80 # number of classes
end2end: True # whether to use end-to-end mode
reg_max: 1 # DFL bins
scales: # model compound scaling constants, i.e. 'model=yolo26n.yaml' will call yolo26.yaml with scale 'n'
  # [depth, width, max_channels]
  n: [0.50, 0.25, 1024] # summary: 260 layers, 2,572,280 parameters, 2,572,280 gradients, 6.1 GFLOPs
  s: [0.50, 0.50, 1024] # summary: 260 layers, 10,009,784 parameters, 10,009,784 gradients, 22.8 GFLOPs
  m: [0.50, 1.00, 512] # summary: 280 layers, 21,896,248 parameters, 21,896,248 gradients, 75.4 GFLOPs
  l: [1.00, 1.00, 512] # summary: 392 layers, 26,299,704 parameters, 26,299,704 gradients, 93.8 GFLOPs
  x: [1.00, 1.50, 512] # summary: 392 layers, 58,993,368 parameters, 58,993,368 gradients, 209.5 GFLOPs

# YOLO26n backbone
backbone:
  # [from, repeats, module, args]
  - [-1, 1, Conv, [64, 3, 2]] # 0-P1/2
  - [-1, 1, Conv, [128, 3, 2]] # 1-P2/4
  - [-1, 2, C3k2, [256, False, 0.25]] # 2-P2/4
  - [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
  - [-1, 2, C3k2, [512, False, 0.25]] # 4-P3/8
  - [-1, 1, Conv, [512, 3, 2]] # 5-P4/16
  - [-1, 2, C3k2, [512, True]] # 6-P4/16
  - [-1, 1, Conv, [1024, 3, 2]] # 7-P5/32
  - [-1, 2, C3k2, [1024, True]] # 8-P5/32
  - [-1, 1, SPPF, [1024, 5, 3, True]] # 9-P5/32
  - [-1, 2, C2PSA, [1024]] # 10-P5/32

# YOLO26n head
head:
  - [[6, -1], 1, WFU, [512]] # 11-P4/16
  - [-1, 2, C3k2, [512, True]] # 12-P4/16

  - [[4, -1], 1, WFU, [256]] # 13-P3/8
  - [-1, 2, C3k2, [256, True]] # 14-P3/8

  - [-1, 1, Conv, [256, 3, 2]] # 15-P4/16
  - [[-1, 12], 1, Concat, [1]] # 16-P4/16
  - [-1, 2, C3k2, [512, True]] # 17-P4/16

  - [-1, 1, Conv, [512, 3, 2]] # 18-P5/32
  - [[-1, 10], 1, Concat, [1]] # 19-P5/32
  - [-1, 1, C3k2, [1024, True, 0.5, True]] # 20-P5/32

  - [[14, 17, 20], 1, Detect, [nc]] # 21-P3/8,P4/16,P5/32

语义分割

python 复制代码

# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license

# Ultralytics YOLO26 object detection model with P3/8 - P5/32 outputs
# Model docs: https://docs.ultralytics.com/models/yolo26
# Task docs: https://docs.ultralytics.com/tasks/detect

# Parameters
nc: 80 # number of classes
end2end: True # whether to use end-to-end mode
reg_max: 1 # DFL bins
scales: # model compound scaling constants, i.e. 'model=yolo26n.yaml' will call yolo26.yaml with scale 'n'
  # [depth, width, max_channels]
  n: [0.50, 0.25, 1024] # summary: 260 layers, 2,572,280 parameters, 2,572,280 gradients, 6.1 GFLOPs
  s: [0.50, 0.50, 1024] # summary: 260 layers, 10,009,784 parameters, 10,009,784 gradients, 22.8 GFLOPs
  m: [0.50, 1.00, 512] # summary: 280 layers, 21,896,248 parameters, 21,896,248 gradients, 75.4 GFLOPs
  l: [1.00, 1.00, 512] # summary: 392 layers, 26,299,704 parameters, 26,299,704 gradients, 93.8 GFLOPs
  x: [1.00, 1.50, 512] # summary: 392 layers, 58,993,368 parameters, 58,993,368 gradients, 209.5 GFLOPs

# YOLO26n backbone
backbone:
  # [from, repeats, module, args]
  - [-1, 1, Conv, [64, 3, 2]] # 0-P1/2
  - [-1, 1, Conv, [128, 3, 2]] # 1-P2/4
  - [-1, 2, C3k2, [256, False, 0.25]] # 2-P2/4
  - [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
  - [-1, 2, C3k2, [512, False, 0.25]] # 4-P3/8
  - [-1, 1, Conv, [512, 3, 2]] # 5-P4/16
  - [-1, 2, C3k2, [512, True]] # 6-P4/16
  - [-1, 1, Conv, [1024, 3, 2]] # 7-P5/32
  - [-1, 2, C3k2, [1024, True]] # 8-P5/32
  - [-1, 1, SPPF, [1024, 5, 3, True]] # 9-P5/32
  - [-1, 2, C2PSA, [1024]] # 10-P5/32

# YOLO26n head
head:
  - [[6, -1], 1, WFU, [512]] # 11-P4/16
  - [-1, 2, C3k2, [512, True]] # 12-P4/16

  - [[4, -1], 1, WFU, [256]] # 13-P3/8
  - [-1, 2, C3k2, [256, True]] # 14-P3/8

  - [-1, 1, Conv, [256, 3, 2]] # 15-P4/16
  - [[-1, 12], 1, Concat, [1]] # 16-P4/16
  - [-1, 2, C3k2, [512, True]] # 17-P4/16

  - [-1, 1, Conv, [512, 3, 2]] # 18-P5/32
  - [[-1, 10], 1, Concat, [1]] # 19-P5/32
  - [-1, 1, C3k2, [1024, True, 0.5, True]] # 20-P5/32

  - [[14, 17, 20], 1, Segment, [nc, 32, 256]]

旋转目标检测

python 复制代码

# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license

# Ultralytics YOLO26 object detection model with P3/8 - P5/32 outputs
# Model docs: https://docs.ultralytics.com/models/yolo26
# Task docs: https://docs.ultralytics.com/tasks/detect

# Parameters
nc: 80 # number of classes
end2end: True # whether to use end-to-end mode
reg_max: 1 # DFL bins
scales: # model compound scaling constants, i.e. 'model=yolo26n.yaml' will call yolo26.yaml with scale 'n'
  # [depth, width, max_channels]
  n: [0.50, 0.25, 1024] # summary: 260 layers, 2,572,280 parameters, 2,572,280 gradients, 6.1 GFLOPs
  s: [0.50, 0.50, 1024] # summary: 260 layers, 10,009,784 parameters, 10,009,784 gradients, 22.8 GFLOPs
  m: [0.50, 1.00, 512] # summary: 280 layers, 21,896,248 parameters, 21,896,248 gradients, 75.4 GFLOPs
  l: [1.00, 1.00, 512] # summary: 392 layers, 26,299,704 parameters, 26,299,704 gradients, 93.8 GFLOPs
  x: [1.00, 1.50, 512] # summary: 392 layers, 58,993,368 parameters, 58,993,368 gradients, 209.5 GFLOPs

# YOLO26n backbone
backbone:
  # [from, repeats, module, args]
  - [-1, 1, Conv, [64, 3, 2]] # 0-P1/2
  - [-1, 1, Conv, [128, 3, 2]] # 1-P2/4
  - [-1, 2, C3k2, [256, False, 0.25]] # 2-P2/4
  - [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
  - [-1, 2, C3k2, [512, False, 0.25]] # 4-P3/8
  - [-1, 1, Conv, [512, 3, 2]] # 5-P4/16
  - [-1, 2, C3k2, [512, True]] # 6-P4/16
  - [-1, 1, Conv, [1024, 3, 2]] # 7-P5/32
  - [-1, 2, C3k2, [1024, True]] # 8-P5/32
  - [-1, 1, SPPF, [1024, 5, 3, True]] # 9-P5/32
  - [-1, 2, C2PSA, [1024]] # 10-P5/32

# YOLO26n head
head:
  - [[6, -1], 1, WFU, [512]] # 11-P4/16
  - [-1, 2, C3k2, [512, True]] # 12-P4/16

  - [[4, -1], 1, WFU, [256]] # 13-P3/8
  - [-1, 2, C3k2, [256, True]] # 14-P3/8

  - [-1, 1, Conv, [256, 3, 2]] # 15-P4/16
  - [[-1, 12], 1, Concat, [1]] # 16-P4/16
  - [-1, 2, C3k2, [512, True]] # 17-P4/16

  - [-1, 1, Conv, [512, 3, 2]] # 18-P5/32
  - [[-1, 10], 1, Concat, [1]] # 19-P5/32
  - [-1, 1, C3k2, [1024, True, 0.5, True]] # 20-P5/32

  - [[14, 17, 20], 1, OBB, [nc, 1]]

温馨提示：本文只是对yolo26基础上添加模块，如果要对yolo26 n/l/m/x进行添加则只需要指定对应的depth_multiple 和 width_multiple

python 复制代码

end2end: True # whether to use end-to-end mode
reg_max: 1 # DFL bins
scales: # model compound scaling constants, i.e. 'model=yolo26n.yaml' will call yolo26.yaml with scale 'n'
  # [depth, width, max_channels]
  n: [0.50, 0.25, 1024] # summary: 260 layers, 2,572,280 parameters, 2,572,280 gradients, 6.1 GFLOPs
  s: [0.50, 0.50, 1024] # summary: 260 layers, 10,009,784 parameters, 10,009,784 gradients, 22.8 GFLOPs
  m: [0.50, 1.00, 512] # summary: 280 layers, 21,896,248 parameters, 21,896,248 gradients, 75.4 GFLOPs
  l: [1.00, 1.00, 512] # summary: 392 layers, 26,299,704 parameters, 26,299,704 gradients, 93.8 GFLOPs
  x: [1.00, 1.50, 512] # summary: 392 layers, 58,993,368 parameters, 58,993,368 gradients, 209.5 GFLOPs

2.4 在task.py中进行注册

**关键步骤四：**在parse_model函数中进行注册，添加WFU

先在task.py导入函数

然后在task.py文件下找到parse_model这个函数，如下图，添加WFU

python 复制代码

        elif m in frozenset({WFU}):  
            c1, c2 = [ch[fi] for fi in f], args[0]
            c2 = make_divisible(min(c2, max_channels) * width, 8)  
            args = [c1, c2, *args[1:]]

2.5 执行程序

关键步骤五: 在ultralytics文件中新建train.py，将model的参数路径设置为yolo26_WFU.yaml的路径即可【注意是在外边的Ultralytics下新建train.py】

python 复制代码

from ultralytics import YOLO
import warnings
warnings.filterwarnings('ignore')
from pathlib import Path
 
if __name__ == '__main__':
 
 
    # 加载模型
    model = YOLO("ultralytics/cfg/26/yolo26.yaml")  # 你要选择的模型yaml文件地址
    # Use the model
    results = model.train(data=r"你的数据集的yaml文件地址",
                          epochs=100, batch=16, imgsz=640, workers=4, name=Path(model.cfg).stem)  # 训练模型

🚀运行程序，如果出现下面的内容则说明添加成功🚀

python 复制代码

                   from  n    params  module                                       arguments                     
  0                  -1  1       464  ultralytics.nn.modules.conv.Conv             [3, 16, 3, 2]                 
  1                  -1  1      4672  ultralytics.nn.modules.conv.Conv             [16, 32, 3, 2]                
  2                  -1  1      6640  ultralytics.nn.modules.block.C3k2            [32, 64, 1, False, 0.25]      
  3                  -1  1     36992  ultralytics.nn.modules.conv.Conv             [64, 64, 3, 2]                
  4                  -1  1     26080  ultralytics.nn.modules.block.C3k2            [64, 128, 1, False, 0.25]     
  5                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]              
  6                  -1  1     87040  ultralytics.nn.modules.block.C3k2            [128, 128, 1, True]           
  7                  -1  1    295424  ultralytics.nn.modules.conv.Conv             [128, 256, 3, 2]              
  8                  -1  1    346112  ultralytics.nn.modules.block.C3k2            [256, 256, 1, True]           
  9                  -1  1    164608  ultralytics.nn.modules.block.SPPF            [256, 256, 5, 3, True]        
 10                  -1  1    249728  ultralytics.nn.modules.block.C2PSA           [256, 256, 1]                 
 11             [6, -1]  1    594944  ultralytics.nn.models.WFU.WFU                [[128, 256], 128]             
 12                  -1  1     87040  ultralytics.nn.modules.block.C3k2            [128, 128, 1, True]           
 13             [4, -1]  1    472064  ultralytics.nn.models.WFU.WFU                [[128, 128], 64]              
 14                  -1  1     22016  ultralytics.nn.modules.block.C3k2            [64, 64, 1, True]             
 15                  -1  1     36992  ultralytics.nn.modules.conv.Conv             [64, 64, 3, 2]                
 16            [-1, 12]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           
 17                  -1  1     95232  ultralytics.nn.modules.block.C3k2            [192, 128, 1, True]           
 18                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]              
 19            [-1, 10]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           
 20                  -1  1    463104  ultralytics.nn.modules.block.C3k2            [384, 256, 1, True, 0.5, True]
 21        [14, 17, 20]  1    309656  ultralytics.nn.modules.head.Detect           [80, 1, True, [64, 128, 256]] 
YOLO26_WFU summary: 275 layers, 3,594,232 parameters, 3,586,040 gradients, 7.9 GFLOPs

3. 完整代码分享

++主页侧边++

4. GFLOPs

关于GFLOPs的计算方式可以查看 ：百面算法工程师 | 卷积基础知识------Convolution

未改进的YOLO26n GFLOPs

改进后的GFLOPs

5. 进阶

可以与其他的注意力机制或者损失函数等结合，进一步提升检测效果

6.总结

通过以上的改进方法，我们成功提升了模型的表现。这只是一个开始，未来还有更多优化和技术深挖的空间。在这里，我想隆重向大家推荐我的专栏------<专栏地址： YOLO26改进-论文涨点------点击跳转看所有内容，关注不迷路！>。这个专栏专注于前沿的深度学习技术，特别是目标检测领域的最新进展，不仅包含对YOLO26的深入解析和改进策略，还会定期更新来自各大顶会（如CVPR、NeurIPS等）的论文复现和实战分享。

为什么订阅我的专栏？ ------专栏地址：YOLO26改进-论文涨点------点击跳转看所有内容，关注不迷路！****

前沿技术解读：专栏不仅限于YOLO系列的改进，还会涵盖各类主流与新兴网络的最新研究成果，帮助你紧跟技术潮流。
详尽的实践分享 ：所有内容实践性也极强。每次更新都会附带代码和具体的改进步骤，保证每位读者都能迅速上手。
问题互动与答疑 ：订阅我的专栏后，你将可以随时向我提问，获取及时的答疑。
实时更新，紧跟行业动态：不定期发布来自全球顶会的最新研究方向和复现实验报告，让你时刻走在技术前沿。

专栏适合人群：

对目标检测、YOLO系列网络有深厚兴趣的同学
希望在用YOLO算法写论文的同学
对YOLO算法感兴趣的同学等