<![CDATA[包括YOLOv5、YOLOv7、YOLOv8等模型改进]]>https://blog.csdn.net/qq_38668236zh-cnhttps://blog.csdn.net/<![CDATA[芒果YOLOv12改进108：基于QARep提出多种＜独家原创结构＞：ReNLANQARep, QARep, CSCQARep, C3QARep, C2fQARep等，即插即用提升小目标检测]]>https://blog.csdn.net/qq_38668236/article/details/147313281https://blog.csdn.net/qq_38668236/article/details/147313281qq_38668236Thu, 17 Apr 2025 20:27:50 +0800<![CDATA[芒果YOLOv12改进107：主干Backbone篇：基于MS思想，提出CrossMScaleB核心结构，增强实时目标检测器的多尺度特征表示]]>https://blog.csdn.net/qq_38668236/article/details/147313268https://blog.csdn.net/qq_38668236/article/details/147313268qq_38668236Thu, 17 Apr 2025 20:27:08 +0800<![CDATA[芒果YOLOv12改进106：检测头篇DynamicHead动态检测头：基于DynamicHead，二次原创提出SATNeXt模块，提出全新的动态DynamicSatDetect]]>https://blog.csdn.net/qq_38668236/article/details/147313255https://blog.csdn.net/qq_38668236/article/details/147313255qq_38668236Thu, 17 Apr 2025 20:26:27 +0800<![CDATA[芒果YOLOv12改进105：卷积Conv篇：DO-DConv卷积提高性能涨点，使用over-parameterized卷积层提高CNN性能]]>https://blog.csdn.net/qq_38668236/article/details/147313231https://blog.csdn.net/qq_38668236/article/details/147313231qq_38668236Thu, 17 Apr 2025 20:25:37 +0800<![CDATA[芒果YOLOv12改进104：基于ConvNeXtv2提出多种原创结构CPNConvNeXtv2, CSCConvv2, ReNLANConvv2, C3_Conv2等，即插即用，提升小目标检测]]>https://blog.csdn.net/qq_38668236/article/details/147313221https://blog.csdn.net/qq_38668236/article/details/147313221qq_38668236Thu, 17 Apr 2025 20:25:01 +0800<![CDATA[芒果YOLOv12改进103：即插即用 | 基于BiFormer提出多种原创结构CSCBiF, ReNLANBiF, CPNBiF, C3_Biformer, C2f_Biformer，提升小目标检测]]>https://blog.csdn.net/qq_38668236/article/details/147313211https://blog.csdn.net/qq_38668236/article/details/147313211qq_38668236Thu, 17 Apr 2025 20:24:25 +0800<![CDATA[芒果YOLOv12改进102：即插即用：集中特征金字塔Centralized Feature Pyramid 高效涨点改进]]>https://blog.csdn.net/qq_38668236/article/details/147313190https://blog.csdn.net/qq_38668236/article/details/147313190qq_38668236Thu, 17 Apr 2025 20:23:11 +0800<![CDATA[芒果YOLOv12改进101：即插即用 | 基于CFNet提出多种原创结构CSCFocalNeXt, ReNLANFocalNeXt, CPNFocalNeXt, C3_FocalNeXt]]>https://blog.csdn.net/qq_38668236/article/details/147313181https://blog.csdn.net/qq_38668236/article/details/147313181qq_38668236Thu, 17 Apr 2025 20:22:33 +0800<![CDATA[芒果YOLOv12改进100：主干Backbone篇：基于CVPR顶会论文的SwinV2核心结构，提出独家原创SwinV2TRX、SwinV2TRY、SwinV2TRXZ等新结构]]>https://blog.csdn.net/qq_38668236/article/details/147313156https://blog.csdn.net/qq_38668236/article/details/147313156qq_38668236Thu, 17 Apr 2025 20:21:40 +0800<![CDATA[芒果YOLOv12改进99:检测头Head篇：即插即用｜新颖轻量化非对称多级压缩LADH检测头，原创改进适配YOLOv8高效检测头，检测头新颖性改进，全网首发独一份]]>https://blog.csdn.net/qq_38668236/article/details/147080305https://blog.csdn.net/qq_38668236/article/details/147080305qq_38668236Thu, 17 Apr 2025 20:11:45 +0800<![CDATA[芒果YOLOv12改进98：损失函数IoU篇之KLD Loss：即插即用原创首发更新KLDLoss改进博客，通过 Kullback-Leibler 散度学习用于目标检测的高精度边界框，提升YOLO11]]>https://blog.csdn.net/qq_38668236/article/details/147312968https://blog.csdn.net/qq_38668236/article/details/147312968qq_38668236Thu, 17 Apr 2025 20:08:12 +0800<![CDATA[芒果YOLOv12改进97：提升篇：深度集成版YOLOv12热力图来了！支持自定义数据集训练出来的模型]]>https://blog.csdn.net/qq_38668236/article/details/147312879https://blog.csdn.net/qq_38668236/article/details/147312879qq_38668236Thu, 17 Apr 2025 20:02:52 +0800<![CDATA[芒果YOLOv12改进96：提升篇：YOLOv12使用SAHI切片辅助推理：《SAHI切片辅助推理解析》推理教程（附推理源代码）]]>https://blog.csdn.net/qq_38668236/article/details/147312851https://blog.csdn.net/qq_38668236/article/details/147312851qq_38668236Thu, 17 Apr 2025 20:01:04 +0800<![CDATA[芒果YOLOv12改进95：多任务之目标对象计数篇：目标对象计数，深度集成版来了！支持自定义数据集训练自定义模型]]>https://blog.csdn.net/qq_38668236/article/details/147312742https://blog.csdn.net/qq_38668236/article/details/147312742qq_38668236Thu, 17 Apr 2025 19:53:03 +0800<![CDATA[芒果YOLOv12改进94：特征融合Neck篇之Scale-sequence-feature-fusion 结构：新颖尺度序列特征融合结构，有效地利用所有金字塔特征图之间的相关性，即插即用提升精度性能]]>https://blog.csdn.net/qq_38668236/article/details/147289018https://blog.csdn.net/qq_38668236/article/details/147289018qq_38668236Thu, 17 Apr 2025 19:51:41 +0800<![CDATA[芒果YOLOv12改进93：检测头篇之感知聚合SERDet检测头：简单高效涨点，即插即用｜检测头新颖改进，全网无重复 YOLO11专属打造]]>https://blog.csdn.net/qq_38668236/article/details/147288991https://blog.csdn.net/qq_38668236/article/details/147288991qq_38668236Wed, 16 Apr 2025 23:18:31 +0800<![CDATA[芒果YOLOv12改进92：主干Backbone篇之UniRepLKNet：极简高效，UniRepLKNet作为改进升级版RepLKNet(博客内附源代码)，适用于图像识别，即插即用打破性能瓶颈]]>https://blog.csdn.net/qq_38668236/article/details/147288984https://blog.csdn.net/qq_38668236/article/details/147288984qq_38668236Wed, 16 Apr 2025 23:17:59 +0800<![CDATA[芒果YOLOv12改进91：动态标签分配策略篇之Nanodet：同时集成VFL全新损失，来打造新颖YOLO11检测器]]>https://blog.csdn.net/qq_38668236/article/details/147288946https://blog.csdn.net/qq_38668236/article/details/147288946qq_38668236Wed, 16 Apr 2025 23:15:53 +0800<![CDATA[芒果YOLOv12改进90：检测头篇之ShareSepHead：即插即用｜原创新颖共享Sep检测头，更省参数量，更高效，打造新型YOLOv8检测器，精度高效涨点]]>https://blog.csdn.net/qq_38668236/article/details/147288929https://blog.csdn.net/qq_38668236/article/details/147288929qq_38668236Wed, 16 Apr 2025 23:15:13 +0800<![CDATA[芒果YOLOv12改进89：损失函数IoU篇之Shape-IoU：考虑边界框形状和比例的更准确的指标]]>https://blog.csdn.net/qq_38668236/article/details/147288870https://blog.csdn.net/qq_38668236/article/details/147288870qq_38668236Wed, 16 Apr 2025 23:11:14 +0800