当前位置： 首页 > news >正文

YOLO26改进 | featurefusion |红外小目标检测的自适应多尺度细节保融模块

news 2026/5/14 3:39:58

💡💡💡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DPCF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！

目录

1.论文

2. DPCF代码实现

2.1 将DPCF添加到YOLO26中

2.2 更改init.py文件

2.3 添加yaml文件

2.4 在task.py中进行注册

2.5 执行程序

3. 完整代码分享

4. GFLOPs

5. 进阶

6.总结

1.论文

论文地址：SAMamba: Adaptive State Space Modeling with Hierarchical Vision for Infrared Small Target Detection

官方代码：官方代码仓库点击即可跳转

2. DPCF代码实现

2.1 将DPCF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DPCF.py，粘贴下面代码

import torch import torch.nn as nn import torch.nn.functional as F try: from mamba_ssm import Mamba except Exception as e: pass from ultralytics.nn.modules.conv import Conv class AdaptiveCombiner(nn.Module): def __init__(self): super(AdaptiveCombiner, self).__init__() # 定义可学习参数d，形状与p和i相同，这里假设p和i的形状为(batch_size, channel, w, h) self.d = nn.Parameter(torch.randn(1, 1, 1, 1)) def forward(self, p, i): batch_size, channel, w, h = p.shape # 将self.d扩展为与p和i相同的形状 d = self.d.expand(batch_size, channel, w, h) edge_att = torch.sigmoid(d) return edge_att * p + (1 - edge_att) * i class DPCF(nn.Module): def __init__(self, in_features, out_features) -> None: super().__init__() self.ac = AdaptiveCombiner() self.tail_conv = Conv(in_features[1], out_features) self.conv1x1 = Conv(in_features[0], in_features[1], 1) if in_features[0] != in_features[1] else nn.Identity() def forward(self, input): x_low, x_high = input x_low = self.conv1x1(x_low) image_size = x_high.size(2) if x_high != None: x_high = torch.chunk(x_high, 4, dim=1) if x_low != None: x_low = F.interpolate(x_low, size=[image_size, image_size], mode='bilinear', align_corners=True) x_low = torch.chunk(x_low, 4, dim=1) x0 = self.ac(x_low[0], x_high[0]) x1 = self.ac(x_low[1], x_high[1]) x2 = self.ac(x_low[2], x_high[2]) x3 = self.ac(x_low[3], x_high[3]) x = torch.cat((x0, x1, x2, x3), dim=1) x = self.tail_conv(x) return x

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_Conv_BCN.yaml文件，粘贴下面的内容

  • 目标检测
# 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: - [-1, 1, nn.Upsample, [None, 2, "nearest"]] # 11-P4/16 - [[-1, 6], 1, Concat, [1]] # 12-P4/16 - [-1, 2, C3k2, [512, True]] # 13-P4/16 - [-1, 1, nn.Upsample, [None, 2, "nearest"]] # 14-P3/8 - [[-1, 4], 1, Concat, [1]] # 15-P3/8 - [-1, 2, C3k2, [256, True]] # 16-P3/8 - [[-1, 13], 1, DPCF, [512]] # 17-P4/16 - [-1, 2, C3k2, [512, True]] # 18-P4/16 - [[-1, 10], 1, DPCF, [1024]] # 19-P5/32 - [-1, 1, C3k2, [1024, True, 0.5, True]] # 20-P5/32 - [[16, 18, 20], 1, Detect, [nc]] # 21-P3/8,P4/16,P5/32
  • 语义分割
# 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: - [-1, 1, nn.Upsample, [None, 2, "nearest"]] # 11-P4/16 - [[-1, 6], 1, Concat, [1]] # 12-P4/16 - [-1, 2, C3k2, [512, True]] # 13-P4/16 - [-1, 1, nn.Upsample, [None, 2, "nearest"]] # 14-P3/8 - [[-1, 4], 1, Concat, [1]] # 15-P3/8 - [-1, 2, C3k2, [256, True]] # 16-P3/8 - [[-1, 13], 1, DPCF, [512]] # 17-P4/16 - [-1, 2, C3k2, [512, True]] # 18-P4/16 - [[-1, 10], 1, DPCF, [1024]] # 19-P5/32 - [-1, 1, C3k2, [1024, True, 0.5, True]] # 20-P5/32 - [[16, 18, 20], 1, Segment, [nc, 32, 256]]
  • 旋转目标检测
# 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: - [-1, 1, nn.Upsample, [None, 2, "nearest"]] # 11-P4/16 - [[-1, 6], 1, Concat, [1]] # 12-P4/16 - [-1, 2, C3k2, [512, True]] # 13-P4/16 - [-1, 1, nn.Upsample, [None, 2, "nearest"]] # 14-P3/8 - [[-1, 4], 1, Concat, [1]] # 15-P3/8 - [-1, 2, C3k2, [256, True]] # 16-P3/8 - [[-1, 13], 1, DPCF, [512]] # 17-P4/16 - [-1, 2, C3k2, [512, True]] # 18-P4/16 - [[-1, 10], 1, DPCF, [1024]] # 19-P5/32 - [-1, 1, C3k2, [1024, True, 0.5, True]] # 20-P5/32 - [[16, 18, 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

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DPCF

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DPCF

elif m in frozenset({DPCF}): 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_DPCF.yaml的路径即可 【注意是在外边的Ultralytics下新建train.py】

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🚀

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 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest'] 12 [-1, 6] 1 0 ultralytics.nn.modules.conv.Concat [1] 13 -1 1 119808 ultralytics.nn.modules.block.C3k2 [384, 128, 1, True] 14 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest'] 15 [-1, 4] 1 0 ultralytics.nn.modules.conv.Concat [1] 16 -1 1 34304 ultralytics.nn.modules.block.C3k2 [256, 64, 1, True] 17 [-1, 13] 1 25089 ultralytics.nn.models.DPCF.DPCF [[64, 128], 128] 18 -1 1 87040 ultralytics.nn.modules.block.C3k2 [128, 128, 1, True] 19 [-1, 10] 1 99329 ultralytics.nn.models.DPCF.DPCF [[128, 256], 256] 20 -1 1 430336 ultralytics.nn.modules.block.C3k2 [256, 256, 1, True, 0.5, True] 21 [16, 18, 20] 1 309656 ultralytics.nn.modules.head.Detect [80, 1, True, [64, 128, 256]] YOLO26_DPCF summary: 264 layers, 2,471,034 parameters, 2,471,034 gradients, 6.2 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！

  1. 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。

  2. 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。

  3. 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。

  4. 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

​

http://www.jsqmd.com/news/812820/

相关文章：

  • NS-USBLoader完整指南：Switch文件管理、RCM注入与游戏传输的一站式解决方案
  • 消费电子创新困局：从3D/4K到流媒体，技术如何重塑家庭娱乐体验
  • 还在为外语游戏和视频发愁？这款实时屏幕翻译神器让你秒懂一切！
  • 2026年高频加热机技术解析：立式数控全自动淬火机床、立式淬火机、立柱移动式伺服数控淬火机床、贵金属熔炼小型熔炼机选择指南 - 优质品牌商家
  • League Akari：终极英雄联盟客户端工具箱完全指南
  • NotebookLM无法读取Zotero本地PDF？资深IT架构师拆解4层权限链（含macOS/Windows/Wine三端实测日志）
  • Rust微信SDK实战：构建高性能、类型安全的微信机器人
  • Illustrator-scripts：从机械重复到创意释放的设计自动化革命
  • 2026年5月更新：剖析北京顶尖操场围网工厂安平县陆安丝网制品有限公司的核心优势 - 2026年企业推荐榜
  • 3步完成微信读书笔记同步：Obsidian Weread插件完整指南
  • 2026年4月比较好的户外led大屏广告代理公司价格，上海花旗大厦广告/上海白玉兰广场广告，户外led大屏广告公司哪家好 - 品牌推荐师
  • IC测试插座技术解析与市场应用实践
  • 从外包程序员到大厂技术专家，我是如何实现逆袭的
  • 别再被POI 5.2.2坑了！手把手教你搞定XSSF和HSSF的自定义字体颜色（附完整代码）
  • 基于SpringBoot+Vue的mvc高校办公室行政事务管理系统管理系统设计与实现【Java+MySQL+MyBatis完整源码】
  • 研发税收抵免：驱动创新的经济杠杆与实操指南
  • 2026乐山配镜技术分享：绵阳眼镜、绵阳配眼镜、自贡眼镜、自贡配眼镜、乐山眼镜、南充眼镜、南充配眼镜、巴中配眼镜选择指南 - 优质品牌商家
  • 2026纺织化工原料选型指南：印染化工原料、循环水水处理药剂、日化化工原料、消毒水处理药剂、消泡剂水处理药剂、漂染化工原料选择指南 - 优质品牌商家
  • 嵌入式开发中CHM文件的应用与优化
  • 电子束光刻掩模误差建模与校正技术解析
  • 蜘蛛池引爬原理到底是什么
  • 如何彻底优化Windows右键菜单：ContextMenuManager终极使用教程
  • dotfiles配置管理：模块化设计与自动化部署提升开发效率
  • 2026年餐饮门店装修技术解析与头部服务商盘点：餐饮空间设计/餐饮设计/餐馆装修/餐馆设计/中式餐厅设计/中餐厅设计/选择指南 - 优质品牌商家
  • 5分钟掌握暗黑2存档编辑：免费开源工具d2s-editor完全指南
  • ARM PMUv3性能监控单元与中断控制寄存器详解
  • AI智能体扩展实战：基于MCP协议构建AlterLab工具箱服务器
  • VR文旅大空间|沉浸式体验重塑文旅新场景
  • 运算放大器1 ppm精度设计：误差源分析与选型策略
  • AMD APU异构计算与能效优化技术解析