<![CDATA[魔改工程师的博客]]>https://blog.csdn.net/yolochangeworldzh-cnhttps://blog.csdn.net/<![CDATA[YOLO26改进 - C3k2 | C3k2模块融合FMB模块 (Feature Modulation Block) 通过多层次特征调制优化空间与通道表征]]>https://blog.csdn.net/yolochangeworld/article/details/167918334https://blog.csdn.net/yolochangeworld/article/details/167918334yolochangeworldWed, 14 Oct 2026 17:26:17 +0800<![CDATA[YOLO26改进 - 特征融合 | HS-FPN高效特征金字塔网络，通过选择性特征交互实现参数优化与性能提升]]>https://blog.csdn.net/yolochangeworld/article/details/167918138https://blog.csdn.net/yolochangeworld/article/details/167918138yolochangeworldWed, 14 Oct 2026 17:18:29 +0800<![CDATA[YOLO26改进 - 特征融合 | GIRAFFEDET之GFPN ：广义特征金字塔网络，高效地融合多尺度特征]]>https://blog.csdn.net/yolochangeworld/article/details/167917805https://blog.csdn.net/yolochangeworld/article/details/167917805yolochangeworldWed, 14 Oct 2026 17:03:56 +0800<![CDATA[YOLO26改进 - 特征融合 | AFPN渐进特征金字塔网络 实现高效多尺度特征融合与检测精度突破]]>https://blog.csdn.net/yolochangeworld/article/details/167917456https://blog.csdn.net/yolochangeworld/article/details/167917456yolochangeworldWed, 14 Oct 2026 16:48:18 +0800<![CDATA[YOLO26改进 - 特征融合 | SSFF+TPE+CPAM 协同,ASF-YOLO 三重模块破解密集小目标分割难题]]>https://blog.csdn.net/yolochangeworld/article/details/167917297https://blog.csdn.net/yolochangeworld/article/details/167917297yolochangeworldWed, 14 Oct 2026 16:45:01 +0800<![CDATA[YOLO26改进 - 特征融合 | 引入聚集-分发机制：Gather-and-Distribute 优化特征金字塔信息流与多尺度特征融合效率]]>https://blog.csdn.net/yolochangeworld/article/details/167917121https://blog.csdn.net/yolochangeworld/article/details/167917121yolochangeworldWed, 14 Oct 2026 16:32:31 +0800<![CDATA[YOLO26改进 - 特征融合 | LCA光照交叉注意力：跨尺度交叉引导与强度增强协同建模，强化多尺度特征交互 | CVPR2025]]>https://blog.csdn.net/yolochangeworld/article/details/167899666https://blog.csdn.net/yolochangeworld/article/details/167899666yolochangeworldWed, 14 Oct 2026 11:17:45 +0800<![CDATA[YOLO26改进 - 特征融合 | HFFE分层特征融合编码器：空间与坐标注意力引导跨层双向交互，增强红外小目标多尺度表达 | TGRS2025]]>https://blog.csdn.net/yolochangeworld/article/details/167899409https://blog.csdn.net/yolochangeworld/article/details/167899409yolochangeworldWed, 14 Oct 2026 11:14:59 +0800<![CDATA[YOLO26改进 - 特征融合 | MSAA多尺度注意力聚合模块， 多尺度卷积融合与双通道注意力机制]]>https://blog.csdn.net/yolochangeworld/article/details/167899041https://blog.csdn.net/yolochangeworld/article/details/167899041yolochangeworldWed, 14 Oct 2026 11:10:36 +0800<![CDATA[YOLO26 改进 - 注意力机制 | DSSA双稀疏自注意力，降低计算量，并精准聚焦高信息量特征交互]]>https://blog.csdn.net/yolochangeworld/article/details/167898246https://blog.csdn.net/yolochangeworld/article/details/167898246yolochangeworldWed, 14 Oct 2026 11:01:45 +0800<![CDATA[YOLO26 改进 - 注意力机制 | ShuffleAttn序列洗牌注意力，解决多向序列建模中的通道异构与信息不对齐问题]]>https://blog.csdn.net/yolochangeworld/article/details/167897352https://blog.csdn.net/yolochangeworld/article/details/167897352yolochangeworldWed, 14 Oct 2026 10:53:42 +0800<![CDATA[YOLO26改进 - 注意力机制 | MFCA频率通道注意力(Multi-Frequency Channel Attention)：解决小目标特征信息少、易被噪声干扰的问题]]>https://blog.csdn.net/yolochangeworld/article/details/167896064https://blog.csdn.net/yolochangeworld/article/details/167896064yolochangeworldWed, 14 Oct 2026 10:42:08 +0800<![CDATA[YOLO26改进-Neck | LPRMAlignUpModule：局部像素关系建模对齐上采样，缓解跨尺度融合中的细节损失 | TPAMI2025]]>https://blog.csdn.net/yolochangeworld/article/details/167813931https://blog.csdn.net/yolochangeworld/article/details/167813931yolochangeworldTue, 13 Oct 2026 23:01:55 +0800<![CDATA[YOLO26改进-Neck | CAFM交叉注意力融合模块，融合双分支特征、强化通道与空间互补信息，助力高光谱SCI重建精细恢复 | TIP 2025]]>https://blog.csdn.net/yolochangeworld/article/details/167813920https://blog.csdn.net/yolochangeworld/article/details/167813920yolochangeworldTue, 13 Oct 2026 23:00:17 +0800<![CDATA[YOLO26改进-Neck | FAAFusion傅里叶角度对齐融合模块，校准跨层方向语义、缓解特征融合错位，助力遥感旋转目标检测高精度涨点 | CVPR 2026]]>https://blog.csdn.net/yolochangeworld/article/details/167813904https://blog.csdn.net/yolochangeworld/article/details/167813904yolochangeworldTue, 13 Oct 2026 22:58:39 +0800<![CDATA[YOLO26改进-Neck | HAFFormer双向模态引导交叉注意力融合模块，互补挖掘VIS/IR信息、自适应门控融合，助力可见光-红外目标检测轻量涨点 | Pattern Recognition]]>https://blog.csdn.net/yolochangeworld/article/details/167813892https://blog.csdn.net/yolochangeworld/article/details/167813892yolochangeworldTue, 13 Oct 2026 22:56:44 +0800<![CDATA[YOLO26改进-Neck | MAF模态感知融合模块，分模态强化热目标与纹理细节、统一红外可见光特征表征，助力图像融合兼顾视觉质量与语义一致性 | TCSVT 2026]]>https://blog.csdn.net/yolochangeworld/article/details/167813865https://blog.csdn.net/yolochangeworld/article/details/167813865yolochangeworldTue, 13 Oct 2026 22:54:16 +0800<![CDATA[YOLO26改进 - C2PSA | C2PSA融合DWM-MSA双窗口多尺度自注意力：双尺度窗口协同建模局部信息与长程依赖 | TIP2025]]>https://blog.csdn.net/yolochangeworld/article/details/167813817https://blog.csdn.net/yolochangeworld/article/details/167813817yolochangeworldTue, 13 Oct 2026 22:48:19 +0800<![CDATA[YOLO26改进 - C3k2 | C3k2融合GCNet之金箍棒块GCBlock 重参数化捕获全局依赖]]>https://blog.csdn.net/yolochangeworld/article/details/167813791https://blog.csdn.net/yolochangeworld/article/details/167813791yolochangeworldTue, 13 Oct 2026 22:44:27 +0800<![CDATA[YOLO26改进 - C3k2 | C3k2模块融合高效融合注意力EFA（Efficient Fusion Attention）提升图像分割中对多尺度目标的捕获精度]]>https://blog.csdn.net/yolochangeworld/article/details/167813731https://blog.csdn.net/yolochangeworld/article/details/167813731yolochangeworldTue, 13 Oct 2026 22:36:49 +0800