- \n
- \u76ee\u6807\u7c7b\u522b： [\u20180\u2019, \u20181\u2019]<\/li>\n
- \u4e2d\u6587\u7c7b\u522b：[\u2018\u91ce\u732a\u2019, \u2018\u91ce\u732a\u2019]<\/li>\n
- \u8bad\u7ec3\u96c6：2014 \u5f20<\/li>\n
- \u9a8c\u8bc1\u96c6：582 \u5f20<\/li>\n
- \u6d4b\u8bd5\u96c6：289 \u5f20<\/li>\n
- \u603b\u8ba1：2885 \u5f20<\/li>\n<\/ul>\n
📄 data.yaml \u914d\u7f6e\u4fe1\u606f<\/h3>\n
\u8be5\u6570\u636e\u96c6\u63d0\u4f9b\u4e86data.yaml\u6587\u4ef6，\u5185\u5bb9\u5982\u4e0b：<\/p>\n
train<\/span>:<\/span> ..\/train\/images
\nval<\/span>:<\/span> ..\/valid\/images
\ntest<\/span>:<\/span> ..\/test\/images<\/p>\nnc<\/span>:<\/span> 2<\/span>
\nnames<\/span>:<\/span> [<\/span>'0'<\/span>,<\/span> '1'<\/span>]<\/span><\/p>\n🖼\ufe0f \u6807\u6ce8\u53ef\u89c6\u5316<\/h3>\n
\u6570\u636e\u96c6\u4e0b\u8f7d<\/h4>\n
\u6570\u636e\u96c6\u4e0b\u8f7d：\u2b07\ufe0f\u2b07\ufe0f\u2b07\ufe0f \u70b9\u51fb\u4e0b\u8f7d<\/h5>\n
![\"\u6807\u6ce8\u56fe1\"](\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260909134902-6aa163ce81dae.jpg\")
<\/p>\n![\"\u6807\u6ce8\u56fe2\"](\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260909134902-6aa163ced03a5.jpg\")
<\/p>\n![\"\u6807\u6ce8\u56fe3\"](\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260909134903-6aa163cf0135d.jpg\")
<\/p>\n\u6570\u636e\u96c6\u8be6\u7ec6\u4ecb\u7ecd<\/h3>\n
YOLO\u91ce\u5916\u6797\u5730\u91ce\u732a\u76ee\u6807\u68c0\u6d4b\u6570\u636e\u96c6\u662f\u4e00\u4e2a\u4e13\u6ce8\u4e8e\u81ea\u7136\u751f\u6001\u73af\u5883\u4e2d\u91ce\u751f\u52a8\u7269\u76d1\u6d4b\u7684\u89c6\u89c9\u6570\u636e\u96c6，\u5176\u540d\u79f0\u548c\u7c7b\u522b\u4fe1\u606f\u8868\u660e\u8be5\u6570\u636e\u96c6\u4e3b\u8981\u7528\u4e8e\u8bc6\u522b\u548c\u8ffd\u8e2a\u91ce\u751f\u73af\u5883\u4e2d\u51fa\u73b0\u7684\u91ce\u732a\u3002\u7ed3\u5408\u201c\u91ce\u5916\u6797\u5730\u201d\u8fd9\u4e00\u573a\u666f\u7279\u5f81，\u53ef\u4ee5\u63a8\u6d4b\u8be5\u6570\u636e\u96c6\u53ef\u80fd\u6765\u6e90\u4e8e\u68ee\u6797\u3001\u56fd\u5bb6\u516c\u56ed\u6216\u751f\u6001\u4fdd\u62a4\u533a\u7b49\u81ea\u7136\u533a\u57df，\u9002\u7528\u4e8e\u91ce\u751f\u52a8\u7269\u4fdd\u62a4\u3001\u751f\u6001\u7814\u7a76\u4ee5\u53ca\u667a\u80fd\u76d1\u63a7\u7cfb\u7edf\u5f00\u53d1\u7b49\u9886\u57df\u3002\u6b64\u7c7b\u6570\u636e\u96c6\u4e3a\u6784\u5efa\u81ea\u52a8\u5316\u91ce\u751f\u52a8\u7269\u76d1\u6d4b\u7cfb\u7edf\u63d0\u4f9b\u4e86\u91cd\u8981\u7684\u57fa\u7840\u652f\u6301，\u5c24\u5176\u5728\u51cf\u5c11\u4eba\u5de5\u5de1\u62a4\u6210\u672c\u548c\u63d0\u5347\u76d1\u6d4b\u6548\u7387\u65b9\u9762\u5177\u6709\u663e\u8457\u4ef7\u503c\u3002<\/p>\n
\u8be5\u6570\u636e\u96c6\u5305\u542b2885\u5f20\u9ad8\u8d28\u91cf\u56fe\u50cf，\u8986\u76d6\u4e86\u591a\u79cd\u91ce\u5916\u6797\u5730\u73af\u5883\u4e0b\u7684\u91ce\u732a\u5b9e\u4f8b，\u5305\u542b2\u4e2a\u7c7b\u522b，\u867d\u7136\u672a\u63d0\u4f9b\u5177\u4f53\u7684\u4e2d\u6587\u548c\u82f1\u6587\u7c7b\u522b\u540d\u79f0，\u4f46\u6839\u636e\u5e38\u89c1\u7684\u76ee\u6807\u68c0\u6d4b\u4efb\u52a1\u8bbe\u5b9a，\u63a8\u6d4b\u5176\u4e2d\u4e00\u4e2a\u7c7b\u522b\u4e3a\u201c\u91ce\u732a\u201d，\u53e6\u4e00\u4e2a\u53ef\u80fd\u662f\u201c\u80cc\u666f\u201d\u6216\u201c\u5176\u4ed6\u52a8\u7269\u201d\u3002\u6570\u636e\u96c6\u7684\u7c7b\u522b\u8bbe\u8ba1\u8f83\u4e3a\u7b80\u6d01，\u805a\u7126\u4e8e\u5355\u4e00\u7269\u79cd\u7684\u68c0\u6d4b\u4efb\u52a1，\u6709\u52a9\u4e8e\u63d0\u9ad8\u6a21\u578b\u5728\u7279\u5b9a\u573a\u666f\u4e0b\u7684\u6cdb\u5316\u80fd\u529b\u548c\u8bc6\u522b\u7cbe\u5ea6\u3002\u56fe\u50cf\u7684\u591a\u6837\u6027\u4e0e\u5b9e\u9645\u5e94\u7528\u573a\u666f\u76f8\u5339\u914d，\u80fd\u591f\u6709\u6548\u53cd\u6620\u771f\u5b9e\u73af\u5883\u4e2d\u7684\u5149\u7167\u3001\u906e\u6321\u548c\u59ff\u6001\u53d8\u5316\u7b49\u590d\u6742\u56e0\u7d20\u3002<\/p>\n
\u5728\u6807\u6ce8\u89c4\u8303\u65b9\u9762，\u8be5\u6570\u636e\u96c6\u91c7\u7528\u4e86\u6807\u51c6\u7684\u76ee\u6807\u68c0\u6d4b\u6807\u6ce8\u683c\u5f0f，\u6bcf\u5f20\u56fe\u7247\u5747\u914d\u6709\u7cbe\u786e\u7684\u8fb9\u754c\u6846\u6807\u6ce8，\u786e\u4fdd\u4e86\u6a21\u578b\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u5bf9\u76ee\u6807\u4f4d\u7f6e\u548c\u5c3a\u5ea6\u7684\u51c6\u786e\u5b66\u4e60\u3002\u8fd9\u79cd\u9ad8\u7cbe\u5ea6\u7684\u6807\u6ce8\u65b9\u5f0f\u63d0\u5347\u4e86\u6570\u636e\u96c6\u7684\u8bad\u7ec3\u4ef7\u503c，\u4f7f\u5176\u6210\u4e3a\u8bad\u7ec3\u548c\u9a8c\u8bc1\u57fa\u4e8eYOLO\u7cfb\u5217\u7b97\u6cd5\u7684\u91ce\u732a\u68c0\u6d4b\u6a21\u578b\u7684\u7406\u60f3\u9009\u62e9\u3002\u540c\u65f6，\u6570\u636e\u96c6\u7684\u89c4\u6a21\u9002\u4e2d，\u65e2\u9002\u5408\u5c0f\u6837\u672c\u5b66\u4e60\u5b9e\u9a8c，\u4e5f\u53ef\u4f5c\u4e3a\u66f4\u5927\u89c4\u6a21\u6570\u636e\u96c6\u7684\u8865\u5145，\u5177\u5907\u826f\u597d\u7684\u6269\u5c55\u6027\u548c\u5b9e\u7528\u6027\u3002<\/p>\n
\u5728\u5178\u578b\u5e94\u7528\u65b9\u5411\u4e0a，\u8be5\u6570\u636e\u96c6\u53ef\u5e7f\u6cdb\u7528\u4e8e\u91ce\u751f\u52a8\u7269\u76d1\u6d4b\u7cfb\u7edf\u3001\u751f\u6001\u7814\u7a76\u8f85\u52a9\u5de5\u5177\u4ee5\u53ca\u667a\u80fd\u5b89\u9632\u9886\u57df\u7684\u52a8\u7269\u5165\u4fb5\u68c0\u6d4b\u7b49\u573a\u666f\u3002\u5bf9\u4e8e\u5b9e\u9645\u843d\u5730，\u5efa\u8bae\u7ed3\u5408\u591a\u4f20\u611f\u5668\u878d\u5408\u6280\u672f（\u5982\u7ea2\u5916\u76f8\u673a\u3001\u58f0\u5b66\u8bbe\u5907）\u63d0\u5347\u68c0\u6d4b\u7cfb\u7edf\u7684\u9c81\u68d2\u6027，\u5e76\u901a\u8fc7\u6301\u7eed\u7684\u6570\u636e\u91c7\u96c6\u548c\u6a21\u578b\u8fed\u4ee3\u4f18\u5316\u68c0\u6d4b\u6027\u80fd\u3002\u6b64\u5916，\u9488\u5bf9\u4e0d\u540c\u5730\u7406\u533a\u57df\u7684\u91ce\u732a\u79cd\u7fa4\u7279\u5f81，\u53ef\u8fdb\u4e00\u6b65\u6269\u5c55\u6570\u636e\u96c6\u7684\u5730\u57df\u8986\u76d6\u8303\u56f4，\u4ee5\u589e\u5f3a\u6a21\u578b\u7684\u8de8\u533a\u57df\u9002\u5e94\u80fd\u529b\u3002<\/p>\n
YOLO26 \u76ee\u6807\u68c0\u6d4b\u7b97\u6cd5\u539f\u7406<\/h3>\n
\u4ece\u5de5\u7a0b\u89d2\u5ea6\u770b，YOLO26 \u6700\u6709\u4ef7\u503c\u7684\u6539\u8fdb\u5c31\u662f\u5e72\u6389\u4e86 NMS \u540e\u5904\u7406，\u8fd9\u5bf9\u90e8\u7f72\u6765\u8bf4\u7701\u4e86\u592a\u591a\u4e8b\u3002<\/p>\n
\u7aef\u5230\u7aef\u65e0 NMS \u63a8\u7406：\u4f20\u7edf YOLO \u6a21\u578b\u5728\u63a8\u7406\u540e\u9700\u8981\u6267\u884c\u975e\u6781\u5927\u503c\u6291\u5236（NMS）\u6765\u53bb\u9664\u5197\u4f59\u68c0\u6d4b\u6846，\u8fd9\u589e\u52a0\u4e86\u540e\u5904\u7406\u5ef6\u8fdf\u548c\u90e8\u7f72\u590d\u6742\u5ea6\u3002YOLO26 \u9ed8\u8ba4\u91c7\u7528\u4e00\u5bf9\u4e00（one-to-one）\u68c0\u6d4b\u5934，\u76f4\u63a5\u8f93\u51fa\u6bcf\u5f20\u56fe\u50cf\u6700\u591a 300 \u4e2a\u68c0\u6d4b\u7ed3\u679c (N, 300, 6)，\u5b8c\u5168\u7701\u53bb NMS \u6b65\u9aa4，\u63a8\u7406\u6d41\u6c34\u7ebf\u5927\u5e45\u7b80\u5316\u3002\u5728 CPU \u4e0a\u7684 ONNX \u63a8\u7406\u901f\u5ea6\u76f8\u6bd4 YOLO11n \u63d0\u5347\u9ad8\u8fbe 43%\u3002<\/p>\n
\u65e0 DFL \u56de\u5f52：YOLO26 \u79fb\u9664\u4e86\u5206\u5e03\u7126\u70b9\u635f\u5931（Distribution Focal Loss, DFL）\u7ed3\u6784，\u68c0\u6d4b\u5934\u66f4\u52a0\u8f7b\u91cf\u3002\u8fb9\u754c\u6846\u56de\u5f52\u4e0d\u518d\u53d7\u9650\u4e8e\u9884\u8bbe\u7684\u79bb\u6563\u533a\u95f4，\u65e2\u964d\u4f4e\u4e86\u6a21\u578b\u590d\u6742\u5ea6\u53c8\u4fdd\u6301\u4e86\u540c\u7b49\u751a\u81f3\u66f4\u4f18\u7684\u56de\u5f52\u7cbe\u5ea6\u3002<\/p>\n
Progressive Loss \u4e0e STAL：\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u901a\u8fc7\u6e10\u8fdb\u635f\u5931（Progressive Loss）\u5c06\u76d1\u7763\u4fe1\u53f7\u9010\u6b65\u4ece\u8f85\u52a9\u5934\u90e8\u8f6c\u79fb\u5230\u63a8\u7406\u9636\u6bb5\u7684\u4e3b\u5934\u90e8\u3002\u5c0f\u76ee\u6807\u611f\u77e5\u6807\u7b7e\u5206\u914d（STAL）\u7b56\u7565\u7279\u522b\u63d0\u5347\u5c0f\u76ee\u6807\u7684\u6b63\u6837\u672c\u6807\u7b7e\u8986\u76d6\u7387，\u6539\u5584\u5c0f\u7269\u4f53\u68c0\u6d4b\u6027\u80fd\u3002<\/p>\n
MuSGD \u6df7\u5408\u4f18\u5316\u5668：\u521b\u65b0\u6027\u5730\u5c06\u5927\u8bed\u8a00\u6a21\u578b\u5e38\u7528\u7684 Muon \u4f18\u5316\u5668\u4e0e SGD \u7ed3\u5408，\u5b9e\u73b0\u66f4\u7a33\u5b9a\u9ad8\u6548\u7684\u8bad\u7ec3\u6536\u655b\u3002\u76f8\u6bd4\u7eaf SGD，MuSGD \u5728\u5927 batch \u8bad\u7ec3\u65f6\u6536\u655b\u66f4\u5feb\u4e14\u4e0d\u6613\u9707\u8361\u3002<\/p>\n
\u6a21\u578b\u5c3a\u5ea6\u4e0e COCO \u6027\u80fd：YOLO26 \u63d0\u4f9b N\/S\/M\/L\/X \u4e94\u79cd\u5c3a\u5ea6：<\/p>\n
\n
\u6a21\u578bmAP50-95CPU ONNX(ms)T4 TensorRT(ms)\u53c2\u6570\u91cf<\/tr>\n \n\n YOLO26n<\/td>\n 40.9<\/td>\n 38.9<\/td>\n 1.7<\/td>\n 2.4M<\/td>\n<\/tr>\n \n YOLO26s<\/td>\n 48.6<\/td>\n 87.2<\/td>\n 2.5<\/td>\n 9.5M<\/td>\n<\/tr>\n \n YOLO26m<\/td>\n 53.1<\/td>\n 220.0<\/td>\n 4.7<\/td>\n 20.4M<\/td>\n<\/tr>\n \n YOLO26l<\/td>\n 55.0<\/td>\n 286.2<\/td>\n 6.2<\/td>\n 24.8M<\/td>\n<\/tr>\n \n YOLO26x<\/td>\n 57.5<\/td>\n 525.8<\/td>\n 11.8<\/td>\n 55.7M<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n Nano \u7248\u672c\u53c2\u6570\u91cf\u4ec5 2.4M，CPU \u63a8\u7406\u901f\u5ea6\u6bd4 YOLO11n \u5feb 43%，\u9002\u5408\u8fb9\u7f18\u8bbe\u5907\u548c\u5c0f\u6570\u636e\u96c6；X \u7248\u672c mAP \u8fbe 57.5，\u9002\u7528\u4e8e\u5bf9\u51c6\u786e\u7387\u8981\u6c42\u6781\u9ad8\u7684\u573a\u666f\u3002<\/p>\n
\u53cc\u5934\u67b6\u6784\u8bbe\u8ba1：YOLO26 \u68c0\u6d4b\u6a21\u578b\u5185\u7f6e\u4e24\u4e2a\u68c0\u6d4b\u5934\u2014\u2014\u4e00\u5bf9\u4e00（\u9ed8\u8ba4，\u65e0 NMS，\u8f93\u51fa 300 \u4e2a\u68c0\u6d4b\u6846）\u548c\u4e00\u5bf9\u591a（\u9700 NMS，\u8f93\u51fa 8400 \u4e2a\u5019\u9009\u6846）\u3002\u8bad\u7ec3\u65f6\u4e00\u5bf9\u591a\u5934\u4f5c\u4e3a\u8f85\u52a9\u76d1\u7763\u4fe1\u53f7\u52a0\u901f\u6536\u655b，\u63a8\u7406\u65f6\u9ed8\u8ba4\u4f7f\u7528\u4e00\u5bf9\u4e00\u5934\u5b9e\u73b0\u7aef\u5230\u7aef\u8f93\u51fa\u3002<\/p>\n
YOLO26 \u652f\u6301\u5168\u90e8\u4e03\u5927\u89c6\u89c9\u4efb\u52a1：\u68c0\u6d4b\u3001\u5b9e\u4f8b\u5206\u5272\u3001\u8bed\u4e49\u5206\u5272\u3001\u6df1\u5ea6\u4f30\u8ba1\u3001\u5206\u7c7b\u3001\u59ff\u6001\u4f30\u8ba1\u548c\u5b9a\u5411\u8fb9\u754c\u6846\u68c0\u6d4b（OBB），\u4e00\u4e2a\u6846\u67b6\u8986\u76d6\u4ece 2D \u68c0\u6d4b\u5230 3D \u611f\u77e5\u7684\u5b8c\u6574\u9700\u6c42\u3002<\/p>\n
\u5173\u952e\u8bad\u7ec3\u53c2\u6570\u914d\u7f6e\u8be6\u89e3<\/h3>\n
YOLO \u7684\u8d85\u53c2\u8bbe\u7f6e\u6709\u4e00\u5b9a\u89c4\u5f8b\u2014\u2014\u4e0d\u540c\u89c4\u6a21\u7684\u6570\u636e\u96c6\u9002\u7528\u7684\u53c2\u6570\u7ec4\u5408\u4e0d\u540c\u3002\u4ee5\u4e0b\u8868\u683c\u662f\u9002\u914d\u672c\u6570\u636e\u96c6\u7684\u5efa\u8bae\u914d\u7f6e\u3002<\/p>\n
\n
\u53c2\u6570\u9ed8\u8ba4\u503c\u672c\u9879\u76ee\u8bbe\u7f6e\u542b\u4e49<\/tr>\n \n\n model<\/td>\n –<\/td>\n yolo26n.pt<\/td>\n \u9884\u8bad\u7ec3\u6743\u91cd，n\/s\/m\/l\/x \u4e94\u79cd\u5c3a\u5ea6<\/td>\n<\/tr>\n \n epochs<\/td>\n 100<\/td>\n 150<\/td>\n \u8bad\u7ec3\u603b\u8f6e\u6570<\/td>\n<\/tr>\n \n imgsz<\/td>\n 640<\/td>\n 416<\/td>\n \u8f93\u5165\u56fe\u50cf\u5c3a\u5bf8<\/td>\n<\/tr>\n \n batch<\/td>\n 16<\/td>\n 64<\/td>\n \u6279\u6b21\u5927\u5c0f，\u6839\u636e\u663e\u5b58\u8c03\u6574<\/td>\n<\/tr>\n \n lr0<\/td>\n 0.01<\/td>\n 0.02<\/td>\n \u521d\u59cb\u5b66\u4e60\u7387<\/td>\n<\/tr>\n \n lrf<\/td>\n 0.01<\/td>\n 0.01<\/td>\n \u6700\u7ec8\u5b66\u4e60\u7387\u56e0\u5b50（lr0 \u00d7 lrf）<\/td>\n<\/tr>\n \n momentum<\/td>\n 0.937<\/td>\n 0.937<\/td>\n SGD \u52a8\u91cf<\/td>\n<\/tr>\n \n weight_decay<\/td>\n 0.0005<\/td>\n 0.0005<\/td>\n \u6743\u91cd\u8870\u51cf（L2 \u6b63\u5219\u5316）<\/td>\n<\/tr>\n \n warmup_epochs<\/td>\n 3.0<\/td>\n 3.0<\/td>\n \u9884\u70ed\u8f6e\u6570<\/td>\n<\/tr>\n \n warmup_momentum<\/td>\n 0.8<\/td>\n 0.8<\/td>\n \u9884\u70ed\u671f\u95f4\u52a8\u91cf\u521d\u59cb\u503c<\/td>\n<\/tr>\n \n box<\/td>\n 7.5<\/td>\n 7.5<\/td>\n \u8fb9\u754c\u6846\u56de\u5f52\u635f\u5931\u6743\u91cd<\/td>\n<\/tr>\n \n cls<\/td>\n 0.5<\/td>\n 0.5<\/td>\n \u5206\u7c7b\u635f\u5931\u6743\u91cd<\/td>\n<\/tr>\n \n dfl<\/td>\n 1.5<\/td>\n 1.5<\/td>\n DFL \u635f\u5931\u6743\u91cd（YOLO26 \u53ef\u7f6e 0）<\/td>\n<\/tr>\n \n patience<\/td>\n 100<\/td>\n -1<\/td>\n \u65e9\u505c\u8f6e\u6570<\/td>\n<\/tr>\n \n cos_lr<\/td>\n False<\/td>\n True<\/td>\n \u4f59\u5f26\u5b66\u4e60\u7387\u8870\u51cf<\/td>\n<\/tr>\n \n close_mosaic<\/td>\n 10<\/td>\n 10<\/td>\n \u6700\u540e N \u8f6e\u5173\u95ed Mosaic \u589e\u5f3a<\/td>\n<\/tr>\n \n optimizer<\/td>\n auto<\/td>\n SGD<\/td>\n \u4f18\u5316\u5668\u9009\u62e9<\/td>\n<\/tr>\n \n amp<\/td>\n True<\/td>\n True<\/td>\n \u81ea\u52a8\u6df7\u5408\u7cbe\u5ea6\u8bad\u7ec3<\/td>\n<\/tr>\n \n freeze<\/td>\n None<\/td>\n 0<\/td>\n \u51bb\u7ed3\u524d N \u5c42\u9aa8\u5e72\u7f51\u7edc\u53c2\u6570<\/td>\n<\/tr>\n \n dropout<\/td>\n 0.0<\/td>\n 0.1<\/td>\n \u5206\u7c7b\u5934 Dropout \u7387，\u9632\u8fc7\u62df\u5408<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n \u635f\u5931\u6743\u91cd\u7684\u8c03\u53c2\u7ecf\u9a8c：\u5982\u679c\u4f60\u53d1\u73b0\u6a21\u578b\u603b\u662f\u628a\u80cc\u666f\u8bef\u68c0\u4e3a\u76ee\u6807（FP \u9ad8），\u4f18\u5148\u964d cls \u6743\u91cd；\u5982\u679c\u6846\u7684\u4f4d\u7f6e\u5bf9\u4f46\u5c3a\u5bf8\u504f\u5dee\u5927，\u4f18\u5148\u8c03 box \u6743\u91cd\u3002<\/p>\n
\u5b66\u4e60\u7387\u8c03\u53c2\u7ecf\u9a8c：<\/p>\n
- \n
- \u8bad\u7ec3\u521d\u671f loss \u4e0d\u4e0b\u964d \u2192 \u8c03\u5927 lr0 \u6216\u589e\u52a0 warmup_epochs<\/li>\n
- \u8bad\u7ec3\u540e\u671f loss \u9707\u8361 \u2192 \u964d\u4f4e lr0，\u589e\u5927 weight_decay<\/li>\n
- \u9a8c\u8bc1\u96c6 loss \u5148\u964d\u540e\u5347（\u8fc7\u62df\u5408）\u2192 \u589e\u5927 dropout\u3001weight_decay \u6216\u51cf\u5c0f epochs<\/li>\n
- \u5c0f\u6570\u636e\u96c6（<100\u5f20）\u2192 lr0=0.001, weight_decay=5e-4, dropout=0.2<\/li>\n
- \u5927\u6570\u636e\u96c6（>1000\u5f20）\u2192 lr0=0.01, weight_decay=5e-4, batch=32+<\/li>\n<\/ul>\n
\u8bad\u7ec3\u53ef\u89c6\u5316\u4e0eTensorBoard：YOLO26 \u9ed8\u8ba4\u5f00\u542f TensorBoard \u65e5\u5fd7\u8bb0\u5f55，\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u53ef\u5b9e\u65f6\u67e5\u770b loss \u66f2\u7ebf\u548c\u6307\u6807\u53d8\u5316\u3002\u5728\u547d\u4ee4\u884c\u6267\u884c tensorboard –logdir runs\/detect\/train \u5373\u53ef\u6253\u5f00 Web \u770b\u677f，\u6bcf\u8f6e\u8bad\u7ec3\u5b8c\u6210\u540e\u81ea\u52a8\u5237\u65b0\u3002\u8bad\u7ec3\u7ed3\u675f\u540e results.png \u751f\u6210\u7684\u516d\u5408\u4e00\u66f2\u7ebf\u56fe\u662f\u6700\u76f4\u89c2\u7684\u8bca\u65ad\u5de5\u5177\u2014\u2014\u4e00\u773c\u5c31\u80fd\u770b\u51fa\u662f\u5426\u8fc7\u62df\u5408\u3001\u662f\u5426\u6536\u655b\u3001\u54ea\u4e2a loss \u5728\u9707\u8361\u3002<\/p>\n
\u65ad\u70b9\u7eed\u8bad\u673a\u5236：\u5982\u679c\u8bad\u7ec3\u610f\u5916\u4e2d\u65ad（\u65ad\u7535\u3001OOM \u7b49），\u4e0d\u7528\u4ece\u5934\u5f00\u59cb\u3002YOLO \u6bcf\u8f6e\u4fdd\u5b58 last.pt，\u6062\u590d\u8bad\u7ec3\u65f6\u6307\u5b9a model=last.pt \u5e76\u8bbe\u7f6e resume=True，\u5b66\u4e60\u7387\u548c\u4f18\u5316\u5668\u72b6\u6001\u90fd\u4f1a\u4ece\u65ad\u70b9\u6062\u590d，\u8bad\u7ec3\u66f2\u7ebf\u4e5f\u548c\u8fde\u7eed\u8bad\u7ec3\u4e00\u6837\u5e73\u6ed1\u3002<\/p>\n
\u6a21\u578b\u8bc4\u4f30\u6307\u6807\u89e3\u8bfb<\/h3>\n
\u8bad\u7ec3\u5b8c\u6210\u540e\u9700\u8981\u79d1\u5b66\u8bc4\u4f30\u6a21\u578b\u6027\u80fd\u3002YOLO26 \u8bad\u7ec3\u5668\u81ea\u52a8\u8f93\u51fa\u4ee5\u4e0b\u6838\u5fc3\u6307\u6807\u3002<\/p>\n
\u8bad\u7ec3\u5b8c\u6210\u540e，YOLO26 \u8bad\u7ec3\u5668\u81ea\u52a8\u8f93\u51fa\u4ee5\u4e0b\u6307\u6807：<\/p>\n
mAP50（mAP@0.5）：IoU \u9608\u503c\u4e3a 0.5 \u65f6\u7684\u5e73\u5747\u7cbe\u5ea6\u5747\u503c，\u53cd\u6620\u6a21\u578b\u7c97\u7c92\u5ea6\u5b9a\u4f4d\u80fd\u529b\u3002mAP50 \u2265 0.90 \u8868\u793a\u5b9a\u4f4d\u6548\u679c\u5f88\u597d\u3002<\/p>\n
mAP50-95（mAP@0.5:0.95）：IoU \u4ece 0.5 \u5230 0.95 \u5171 10 \u4e2a\u9608\u503c\u4e0a\u53d6\u5e73\u5747\u3002\u66f4\u4e25\u82db，\u53cd\u6620\u7cbe\u786e\u5b9a\u4f4d\u80fd\u529b，\u662f\u66f4\u5177\u4ee3\u8868\u6027\u7684\u7efc\u5408\u8bc4\u4ef7\u6307\u6807\u3002<\/p>\n
Precision（\u7cbe\u786e\u7387）：\u68c0\u6d4b\u7ed3\u679c\u4e2d\u6b63\u786e\u68c0\u6d4b\u7684\u6bd4\u4f8b\u3002\u9ad8 Precision \u610f\u5473\u7740\u865a\u8b66\u5c11\u3002<\/p>\n
Recall（\u53ec\u56de\u7387）：\u771f\u5b9e\u76ee\u6807\u4e2d\u88ab\u6210\u529f\u68c0\u6d4b\u5230\u7684\u6bd4\u4f8b\u3002\u9ad8 Recall \u610f\u5473\u7740\u6f0f\u68c0\u5c11\u3002<\/p>\n
F1 Score：Precision \u548c Recall \u7684\u8c03\u548c\u5e73\u5747\u6570 = 2\u00d7P\u00d7R\/(P+R)，\u7efc\u5408\u8861\u91cf\u6a21\u578b\u5e73\u8861\u6027\u3002<\/p>\n
Confusion Matrix（\u6df7\u6dc6\u77e9\u9635）：N\u00d7N \u77e9\u9635，\u5bf9\u89d2\u7ebf\u4e3a\u6b63\u786e\u5206\u7c7b，\u975e\u5bf9\u89d2\u7ebf\u4e3a\u8bef\u5206\u7c7b，\u53ef\u5b9a\u4f4d\u7c7b\u522b\u95f4\u6df7\u6dc6\u60c5\u51b5\u3002<\/p>\n
\u8bad\u7ec3\u66f2\u7ebf\u5206\u6790：<\/p>\n
- \n
- train\/box_loss \u5355\u8c03\u4e0b\u964d \u2192 \u8fb9\u754c\u6846\u56de\u5f52\u6b63\u5e38\u6536\u655b<\/li>\n
- val\/box_loss \u5148\u964d\u540e\u5347 \u2192 \u8fc7\u62df\u5408\u98ce\u9669，\u5e94\u589e\u5927\u6570\u636e\u589e\u5f3a\u6216\u51cf\u5c0f\u6a21\u578b<\/li>\n
- metrics\/mAP50(B) \u589e\u957f\u653e\u7f13\u8d8b\u4e8e\u5e73\u7a33 \u2192 \u6a21\u578b\u63a5\u8fd1\u6536\u655b\u4e0a\u9650<\/li>\n
- train \u6301\u7eed\u964d\u3001val \u6301\u7eed\u5347 \u2192 \u4e25\u91cd\u8fc7\u62df\u5408，\u5efa\u8bae\u51cf\u5c0f\u6a21\u578b\u6216\u589e\u5927 dropout<\/li>\n<\/ul>\n
\u6307\u6807\u4f7f\u7528\u573a\u666f：<\/p>\n
- \n
- \u5b89\u5168\u751f\u4ea7\u573a\u666f（\u5982\u5b89\u5168\u5e3d\u68c0\u6d4b）\u2192 \u4f18\u5148\u4fdd\u8bc1 Recall，\u5b81\u53ef\u591a\u62a5\u4e0d\u80fd\u6f0f\u62a5<\/li>\n
- \u8d28\u68c0\u573a\u666f（\u5982\u7f3a\u9677\u68c0\u6d4b）\u2192 \u4f18\u5148\u4fdd\u8bc1 Precision，\u964d\u4f4e\u8bef\u68c0\u7387\u51cf\u5c11\u4eba\u5de5\u590d\u6838\u91cf<\/li>\n
- \u901a\u7528\u68c0\u6d4b \u2192 \u5747\u8861\u4f18\u5316，\u5173\u6ce8 mAP50-95 \u548c F1 Score<\/li>\n<\/ul>\n
PR \u66f2\u7ebf\u4e0b\u9762\u79ef\u5c31\u662f AP\u3002\u4f46\u66f4\u91cd\u8981\u7684\u662f\u66f2\u7ebf\u5728\u4e0d\u540c Recall \u533a\u95f4\u7684\u8868\u73b0：Recall < 0.5 \u65f6 Precision \u66b4\u8dcc\u8bf4\u660e\u6a21\u578b\u592a\u4fdd\u5b88，Recall > 0.8 \u65f6 Precision \u8fd8\u884c\u8bf4\u660e\u53ec\u56de\u80fd\u529b\u5f3a\u3002<\/p>\n
YOLO \u7cfb\u5217\u7b97\u6cd5\u6f14\u8fdb\u53f2<\/h3>\n
\u56de\u770b YOLO \u8fd9\u5341\u591a\u5e74\u7684\u8fdb\u5316\u53f2，\u4f60\u4f1a\u53d1\u73b0\u76ee\u6807\u68c0\u6d4b\u7684\u6838\u5fc3\u77db\u76fe\u59cb\u7ec8\u6ca1\u53d8\u2014\u2014\u5982\u4f55\u5728\u901f\u5ea6\u548c\u7cbe\u5ea6\u4e4b\u95f4\u53d6\u5f97\u6700\u4f73\u5e73\u8861\u3002\u53ea\u662f\u89e3\u51b3\u8fd9\u4e2a\u77db\u76fe\u7684\u624b\u6bb5\u5728\u4e0d\u65ad\u5347\u7ea7\u3002<\/p>\n
YOLOv1 (2015)：\u5c06\u68c0\u6d4b\u95ee\u9898\u5efa\u6a21\u4e3a\u5355\u4e00\u56de\u5f52\u95ee\u9898，\u76f4\u63a5\u5728\u8f93\u51fa\u5c42\u9884\u6d4b\u8fb9\u754c\u6846\u548c\u7c7b\u522b\u6982\u7387\u3002\u901f\u5ea6\u6781\u5feb\u4f46\u5b9a\u4f4d\u7cbe\u5ea6\u8f83\u5dee\u3002<\/p>\n
YOLOv2\/YOLOv3 (2016-2018)：\u5f15\u5165 anchor box \u673a\u5236\u3001\u7279\u5f81\u91d1\u5b57\u5854（FPN）\u548c\u591a\u5c3a\u5ea6\u8bad\u7ec3\u3002YOLOv3 \u7684 Darknet-53 \u9aa8\u5e72\u7f51\u7edc\u6210\u4e3a\u7ecf\u5178\u67b6\u6784，\u81f3\u4eca\u4ecd\u88ab\u5e7f\u6cdb\u4f7f\u7528\u3002<\/p>\n
YOLOv5 (2020)：Ultralytics \u5728 PyTorch \u4e0a\u7684\u5b9e\u73b0，\u5e26\u6765\u4e86\u5b8c\u5584\u7684\u8bad\u7ec3\u6846\u67b6\u3001\u81ea\u52a8\u951a\u6846\u805a\u7c7b\u548c\u6570\u636e\u589e\u5f3a pipeline，\u5927\u5e45\u964d\u4f4e\u4e86\u4f7f\u7528\u95e8\u69db\u3002<\/p>\n
YOLOv8 (2023)：\u7edf\u4e00\u4e86\u68c0\u6d4b\u3001\u5206\u5272\u3001\u5206\u7c7b\u548c\u5173\u952e\u70b9\u4efb\u52a1\u7684\u6846\u67b6，\u5f15\u5165 C2f \u6a21\u5757\u548c\u65e0\u951a\u6846（anchor-free）\u68c0\u6d4b\u5934\u3002<\/p>\n
YOLOv9 (2024)：\u63d0\u51fa GELAN（\u901a\u7528\u9ad8\u6548\u5c42\u805a\u5408\u7f51\u7edc）\u548c PGI（\u53ef\u7f16\u7a0b\u68af\u5ea6\u4fe1\u606f），\u540c\u7b49\u53c2\u6570\u91cf\u4e0b\u7cbe\u5ea6\u63d0\u5347\u663e\u8457\u3002<\/p>\n
YOLOv10\/YOLOv11 (2024)：YOLOv10 \u9996\u6b21\u63d0\u51fa\u65e0 NMS \u7aef\u5230\u7aef\u68c0\u6d4b\u3002YOLOv11 \u8fdb\u4e00\u6b65\u4f18\u5316\u9aa8\u5e72\u7f51\u7edc\u548c\u8bad\u7ec3\u7b56\u7565，\u5728\u901f\u5ea6\u548c\u7cbe\u5ea6\u4e4b\u95f4\u53d6\u5f97\u65b0\u5e73\u8861\u3002<\/p>\n
YOLO26 (2026)：\u96c6\u5386\u4ee3\u4e4b\u5927\u6210，\u6838\u5fc3\u521b\u65b0\u5305\u62ec：\u2460 \u7aef\u5230\u7aef\u65e0 NMS \u4e00\u5bf9\u4e00\u68c0\u6d4b\u5934 \u2461 \u65e0 DFL \u7684\u7b80\u5316\u56de\u5f52\u5934 \u2462 Progressive Loss \u6e10\u8fdb\u5f0f\u76d1\u7763\u8f6c\u79fb \u2463 STAL \u5c0f\u76ee\u6807\u611f\u77e5\u6807\u7b7e\u5206\u914d \u2464 MuSGD \u6df7\u5408\u4f18\u5316\u5668\u3002<\/p>\n
\u5173\u952e\u8bbe\u8ba1\u6f14\u8fdb\u603b\u7ed3：<\/p>\n
- \n
- Anchor-based \u2192 Anchor-free（v8 \u8d77）<\/li>\n
- C3 \u2192 C2f \u2192 C3k2（\u9aa8\u5e72\u6a21\u5757\u6301\u7eed\u4f18\u5316）<\/li>\n
- \u591a\u4efb\u52a1\u7edf\u4e00\u6846\u67b6（v8 \u8d77）<\/li>\n
- NMS \u540e\u5904\u7406 \u2192 \u65e0 NMS \u7aef\u5230\u7aef（v10\/v26）<\/li>\n
- \u5355\u4e00\u635f\u5931\u51fd\u6570 \u2192 \u591a\u4efb\u52a1\u8054\u5408\u635f\u5931（Box + Cls + DFL，v26 \u53ef\u53bb DFL）<\/li>\n<\/ul>\n
YOLO26 \u8bad\u7ec3\u6b65\u9aa4\u8be6\u89e3<\/h3>\n
\u5199\u4e2a Python \u811a\u672c\u63a7\u5236\u8bad\u7ec3\u6d41\u7a0b，\u6bd4\u547d\u4ee4\u884c\u7075\u6d3b\u4e0d\u5c11<\/p>\n
\u73af\u5883\u51c6\u5907：<\/p>\n
pip install<\/span> ultralytics
\nyolo checks
\npython -c<\/span> "import torch; print(torch.cuda.is_available())"<\/span><\/p>\n\u547d\u4ee4\u884c\u8bad\u7ec3：<\/p>\n
yolo detect train data<\/span>=<\/span>data.yaml model<\/span>=<\/span>yolo26n.pt epochs<\/span>=<\/span>150<\/span> imgsz<\/span>=<\/span>416<\/span> batch<\/span>=<\/span>64<\/span> device<\/span>=<\/span>0<\/span> lr0<\/span>=<\/span>0.02<\/span> patience<\/span>=<\/span>-1 cos_lr<\/span>=<\/span>True close_mosaic<\/span>=<\/span>10<\/span><\/p>\n
Python API \u8bad\u7ec3：<\/p>\n
from<\/span> ultralytics import<\/span> YOLO<\/p>\n
model =<\/span> YOLO(<\/span>"yolo26n.pt"<\/span>)<\/span>
\nresults =<\/span> model.<\/span>train(<\/span>
\n data=<\/span>"data.yaml"<\/span>,<\/span>
\n epochs=<\/span>150<\/span>,<\/span>
\n imgsz=<\/span>416<\/span>,<\/span>
\n batch=<\/span>64<\/span>,<\/span>
\n device=<\/span>0<\/span>,<\/span>
\n lr0=<\/span>0.02<\/span>,<\/span>
\n patience=<\/span>–<\/span>1<\/span>,<\/span>
\n cos_lr=<\/span>True<\/span>,<\/span>
\n close_mosaic=<\/span>10<\/span>,<\/span>
\n augment=<\/span>True<\/span>,<\/span>
\n amp=<\/span>True<\/span>,<\/span>
\n workers=<\/span>8<\/span>,<\/span>
\n)<\/span><\/p>\n\u8bad\u7ec3\u8f93\u51fa\u6587\u4ef6：<\/p>\n
- \n
- runs\/detect\/train\/weights\/best.pt：\u9a8c\u8bc1\u96c6\u4e0a mAP \u6700\u9ad8\u7684\u6743\u91cd<\/li>\n
- runs\/detect\/train\/weights\/last.pt：\u6700\u540e\u4e00\u8f6e\u4fdd\u5b58\u7684\u6743\u91cd<\/li>\n
- runs\/detect\/train\/results.csv：\u6bcf\u8f6e loss \u548c\u6307\u6807\u6570\u636e<\/li>\n
- runs\/detect\/train\/confusion_matrix.png：\u6df7\u6dc6\u77e9\u9635\u56fe<\/li>\n
- runs\/detect\/train\/results.png：\u8bad\u7ec3\u66f2\u7ebf\u56fe<\/li>\n
- runs\/detect\/train\/val_batch*_pred.jpg：\u9a8c\u8bc1\u96c6\u9884\u6d4b\u6548\u679c\u56fe<\/li>\n<\/ul>\n
\u5355\u5f20\u63a8\u7406\u6d4b\u8bd5：<\/p>\n
model =<\/span> YOLO(<\/span>"runs\/detect\/train\/weights\/best.pt"<\/span>)<\/span>
\nresults =<\/span> model(<\/span>"test_image.jpg"<\/span>)<\/span>
\nresults[<\/span>0<\/span>]<\/span>.<\/span>show(<\/span>)<\/span><\/p>\n\u591a GPU \u5206\u5e03\u5f0f\u8bad\u7ec3：<\/p>\n
yolo detect train data<\/span>=<\/span>data.yaml model<\/span>=<\/span>yolo26n.pt device<\/span>=<\/span>0,1<\/span> epochs<\/span>=<\/span>100<\/span><\/p>\n
\u6062\u590d\u4e2d\u65ad\u7684\u8bad\u7ec3：<\/p>\n
yolo detect train model<\/span>=<\/span>path\/to\/last.pt data<\/span>=<\/span>data.yaml resume<\/span>=<\/span>True<\/p>\n
\u90e8\u7f72\u4e0e\u5e94\u7528\u5efa\u8bae<\/h3>\n
\u5927\u89c4\u6a21\u751f\u4ea7\u90e8\u7f72：2885 \u5f20\u8bad\u7ec3\u6570\u636e + YOLO26，\u7cbe\u5ea6\u57fa\u7840\u5df2\u7ecf\u5f88\u597d\u3002\u91cd\u70b9\u662f\u63a8\u7406\u541e\u5410\u548c\u7a33\u5b9a\u6027\u3002<\/p>\n
model.<\/span>export(<\/span>format<\/span>=<\/span>"engine"<\/span>,<\/span> imgsz=<\/span>416<\/span>,<\/span> half=<\/span>True<\/span>,<\/span> batch=<\/span>32<\/span>)<\/span>
\n# TensorRT FP16 + \u5927 batch \u2192 \u5355 GPU \u8f7b\u677e 500+ FPS<\/span><\/p>\n\u751f\u4ea7\u73af\u5883 checklist：\u2460 \u5bfc\u51fa\u65f6\u6307\u5b9a opset=12（ONNX）\/ workspace=4GB（TensorRT）\u2461 \u505a\u4e00\u8f6e int8 \u91cf\u5316\u5bf9\u6bd4，\u786e\u8ba4\u7cbe\u5ea6\u635f\u5931\u53ef\u63a7 \u2462 \u538b\u6d4b\u65f6\u540c\u65f6\u6d4b 1\/4\/8\/16\/32 \u7684 batch \u541e\u5410，\u627e\u5230\u6700\u4f18\u914d\u7f6e \u2463 \u8bbe\u7f6e P99 < 50ms \u7684\u76d1\u63a7\u544a\u8b66\u3002<\/p>\n
\u8bad\u7ec3\u7684\u5751\u4e0e\u586b\u5751\u6307\u5357：\u8bad\u4e86\u8fd9\u4e48\u591a\u6b21 YOLO，\u6211\u53d1\u73b0\u8fd9\u89c4\u6a21\u7684\u8bad\u7ec3\u5bb9\u6613\u51fa\u73b0\u51e0\u4e2a\u5178\u578b\u95ee\u9898\u3002<\/p>\n
\u7b2c\u4e00\u4e2a\u5751\u662f batch size \u548c imgsz \u7684\u914d\u5408\u3002\u4e2d\u7b49\u6570\u636e\u96c6\u4e0b batch=16 \u662f\u4e2a\u8d77\u70b9，\u4f46\u5982\u679c\u663e\u5b58\u591f，batch=32 \u4f1a\u8ba9\u68af\u5ea6\u4f30\u8ba1\u66f4\u7a33\u5b9a\u3002\u4e0d\u8981\u4e3a\u4e86\u7701\u663e\u5b58\u628a imgsz \u964d\u592a\u591a\u2014\u2014416 \u4ee5\u4e0b\u5c0f\u76ee\u6807\u68c0\u6d4b\u5bb9\u6613\u7ffb\u8f66\u3002<\/p>\n
\u7b2c\u4e8c\u4e2a\u5751\u662f\u9a8c\u8bc1\u96c6\u5212\u5206\u3002\u81ea\u52a8 split \u51fa\u6765\u7684\u9a8c\u8bc1\u96c6\u5982\u679c\u4e0d\u505a\u68c0\u67e5，\u53ef\u80fd\u548c\u8bad\u7ec3\u96c6\u6709\u5927\u91cf\u91cd\u590d（\u7279\u522b\u662f\u89c6\u9891\u622a\u5e27\u6570\u636e\u96c6），\u5bfc\u81f4\u9a8c\u8bc1\u6307\u6807\u865a\u9ad8\u3002\u8dd1\u5b8c split \u540e\u968f\u673a\u62bd 10 \u5bf9 train\/val \u6837\u672c\u7528\u8089\u773c\u770b\u770b\u6709\u6ca1\u6709\u9ad8\u5ea6\u76f8\u4f3c\u7684\u3002<\/p>\n
\u7b2c\u4e09\u4e2a\u5751\u662f\u9884\u8bad\u7ec3\u6743\u91cd\u7684\u7248\u672c\u5339\u914d\u3002\u5982\u679c\u7528\u7684\u662f yolo26n.pt \u4f46\u8bad\u7ec3\u4ee3\u7801\u662f\u65e7\u7248 YOLO，\u52a0\u8f7d\u6743\u91cd\u65f6\u53ef\u80fd silently fail（\u90e8\u5206\u5c42\u4e0d\u5339\u914d\u4f46\u4e0d\u62a5\u9519）\u3002\u8bad\u7ec3\u524d\u7528 model.info() \u786e\u8ba4\u6240\u6709\u5c42\u90fd\u6b63\u5e38\u52a0\u8f7d\u4e86\u3002<\/p>\n
\u6301\u7eed\u4f18\u5316\u7684\u56db\u4e2a\u65b9\u5411：\u6a21\u578b\u8bad\u597d\u540e\u4e0d\u662f\u7ec8\u70b9，\u4ee5\u4e0b\u65b9\u5411\u53ef\u4ee5\u6301\u7eed\u63d0\u5347\u6548\u679c\u3002<\/p>\n
\u4e00\u662f\u589e\u91cf\u8bad\u7ec3\u7b56\u7565\u3002\u5982\u679c\u540e\u7eed\u6709\u65b0\u6807\u6ce8\u6570\u636e\u8fdb\u6765，\u522b\u4ece\u5934\u8bad，\u800c\u662f\u5728 best.pt \u57fa\u7840\u4e0a\u7528\u4f4e\u5b66\u4e60\u7387（lr0=0.001）\u518d\u8dd1 20-30 \u8f6e，\u8fd9\u53eb incremental learning，\u65e2\u7701\u65f6\u53c8\u80fd\u4fdd\u6301\u539f\u6709\u80fd\u529b\u3002<\/p>\n
\u4e8c\u662f\u6a21\u578b\u84b8\u998f\u3002\u5982\u679c\u90e8\u7f72\u65f6\u9700\u8981\u5c0f\u6a21\u578b\u4f46\u7cbe\u5ea6\u4e0d\u591f，\u5148\u7528 yolo26x \u8bad\u4e00\u4e2a\u5927\u6a21\u578b，\u518d\u7528\u5b83\u4f5c\u4e3a\u6559\u5e08\u6307\u5bfc yolo26n \u8bad\u7ec3\u3002\u84b8\u998f\u4ee3\u7801 YOLO26 \u539f\u751f\u652f\u6301，\u53c2\u6570 teacher=teacher_model.pt \u4e00\u884c\u914d\u7f6e\u5373\u53ef\u3002<\/p>\n
\u4e09\u662f\u6570\u636e\u98de\u8f6e\u3002\u628a\u6a21\u578b\u90e8\u7f72\u5230\u751f\u4ea7\u7ebf\u540e，\u6536\u96c6\u68c0\u6d4b\u7f6e\u4fe1\u5ea6\u5728 0.3-0.6 \u4e4b\u95f4\u7684"\u4e0d\u786e\u5b9a"\u68c0\u6d4b\u7ed3\u679c，\u4eba\u5de5\u590d\u6838\u540e\u52a0\u5165\u8bad\u7ec3\u96c6\u3002\u8fd9\u79cd\u4e3b\u52a8\u5b66\u4e60（active learning）\u7b56\u7565\u80fd\u7528\u6700\u5c11\u7684\u6807\u6ce8\u6210\u672c\u6362\u53d6\u6700\u5927\u7684\u7cbe\u5ea6\u63d0\u5347\u3002<\/p>\n
\u56db\u662f\u91cf\u5316\u90e8\u7f72\u3002\u5bfc\u51fa TFLite \u6216 TensorRT \u65f6\u5f00\u542f int8 \u91cf\u5316，\u6a21\u578b\u4f53\u79ef\u7f29\u81f3\u56db\u5206\u4e4b\u4e00，\u63a8\u7406\u901f\u5ea6\u5feb 2-3 \u500d，\u7cbe\u5ea6\u635f\u5931\u901a\u5e38\u5728\u4e00\u4e2a mAP \u70b9\u4ee5\u5185\u3002\u4f46\u91cf\u5316\u524d\u52a1\u5fc5\u5728\u76ee\u6807\u786c\u4ef6\u4e0a\u7528\u9a8c\u8bc1\u96c6\u91cd\u65b0\u8bc4\u4f30\u7cbe\u5ea6，\u4e0d\u540c\u82af\u7247\u7684\u91cf\u5316\u8bef\u5dee\u8868\u73b0\u4e0d\u540c\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"
YOLO\u91ce\u5916\u6797\u5730\u91ce\u732a\u76ee\u6807\u68c0\u6d4b\u6570\u636e\u96c6 \u751f\u7269\u591a\u6837\u6027\u76d1\u6d4b：YOLO26 \u52a8\u7269\u76ee\u6807\u68c0\u6d4b\u5b9e\u8df5 📊 \u6570\u636e\u96c6\u57fa\u672c\u4fe1\u606f
\n\u76ee\u6807\u7c7b\u522b： [\u20180\u2019, \u20181\u2019]\u4e2d\u6587\u7c7b\u522b：[\u2018\u91ce\u732a\u2019, \u2018\u91ce\u732a\u2019]\u8bad\u7ec3\u96c6：2014 \u5f20\u9a8c\u8bc1\u96c6：582 \u5f20\u6d4b\u8bd5\u96c6：289 \u5f20\u603b\u8ba1：2885 \u5f20
\n📄 data.yaml \u914d\u7f6e\u4fe1\u606f
\n\u8be5\u6570\u636e\u96c6\u63d0\u4f9b\u4e86data.yaml\u6587\u4ef6，\u5185\u5bb9\u5982\u4e0b&#<\/p>\n","protected":false},"author":2,"featured_media":103085,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[156,1665,5281],"topic":[],"class_list":["post-103088","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-server","tag-yolo","tag-1665","tag-5281"],"yoast_head":"\nYOLO\u91ce\u5916\u6797\u5730\u91ce\u732a\u76ee\u6807\u68c0\u6d4b\u6570\u636e\u96c6-2885\u5f20 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.wsisp.com\/helps\/103088.html\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"YOLO\u91ce\u5916\u6797\u5730\u91ce\u732a\u76ee\u6807\u68c0\u6d4b\u6570\u636e\u96c6-2885\u5f20 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"og:description\" content=\"YOLO\u91ce\u5916\u6797\u5730\u91ce\u732a\u76ee\u6807\u68c0\u6d4b\u6570\u636e\u96c6 \u751f\u7269\u591a\u6837\u6027\u76d1\u6d4b：YOLO26 \u52a8\u7269\u76ee\u6807\u68c0\u6d4b\u5b9e\u8df5 📊 \u6570\u636e\u96c6\u57fa\u672c\u4fe1\u606f \u76ee\u6807\u7c7b\u522b： [\u20180\u2019, \u20181\u2019]\u4e2d\u6587\u7c7b\u522b：[\u2018\u91ce\u732a\u2019, \u2018\u91ce\u732a\u2019]\u8bad\u7ec3\u96c6：2014 \u5f20\u9a8c\u8bc1\u96c6：582 \u5f20\u6d4b\u8bd5\u96c6：289 \u5f20\u603b\u8ba1：2885 \u5f20 📄 data.yaml \u914d\u7f6e\u4fe1\u606f \u8be5\u6570\u636e\u96c6\u63d0\u4f9b\u4e86data.yaml\u6587\u4ef6，\u5185\u5bb9\u5982\u4e0b&#\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.wsisp.com\/helps\/103088.html\" \/>\n<meta property=\"og:site_name\" content=\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"article:published_time\" content=\"2026-09-09T13:49:04+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260909134902-6aa163ce81dae.jpg\" \/>\n<meta name=\"author\" content=\"admin\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u4f5c\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4\" \/>\n\t<meta name=\"twitter:data2\" content=\"5 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/103088.html\",\"url\":\"https:\/\/www.wsisp.com\/helps\/103088.html\",\"name\":\"YOLO\u91ce\u5916\u6797\u5730\u91ce\u732a\u76ee\u6807\u68c0\u6d4b\u6570\u636e\u96c6-2885\u5f20 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"isPartOf\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\"},\"datePublished\":\"2026-09-09T13:49:04+00:00\",\"dateModified\":\"2026-09-09T13:49:04+00:00\",\"author\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\"},\"breadcrumb\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/103088.html#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.wsisp.com\/helps\/103088.html\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/103088.html#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/www.wsisp.com\/helps\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"YOLO\u91ce\u5916\u6797\u5730\u91ce\u732a\u76ee\u6807\u68c0\u6d4b\u6570\u636e\u96c6-2885\u5f20\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\",\"url\":\"https:\/\/www.wsisp.com\/helps\/\",\"name\":\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"description\":\"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\",\"name\":\"admin\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"contentUrl\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"caption\":\"admin\"},\"sameAs\":[\"http:\/\/wp.wsisp.com\"],\"url\":\"https:\/\/www.wsisp.com\/helps\/author\/admin\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"YOLO\u91ce\u5916\u6797\u5730\u91ce\u732a\u76ee\u6807\u68c0\u6d4b\u6570\u636e\u96c6-2885\u5f20 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.wsisp.com\/helps\/103088.html","og_locale":"zh_CN","og_type":"article","og_title":"YOLO\u91ce\u5916\u6797\u5730\u91ce\u732a\u76ee\u6807\u68c0\u6d4b\u6570\u636e\u96c6-2885\u5f20 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","og_description":"YOLO\u91ce\u5916\u6797\u5730\u91ce\u732a\u76ee\u6807\u68c0\u6d4b\u6570\u636e\u96c6 \u751f\u7269\u591a\u6837\u6027\u76d1\u6d4b：YOLO26 \u52a8\u7269\u76ee\u6807\u68c0\u6d4b\u5b9e\u8df5 📊 \u6570\u636e\u96c6\u57fa\u672c\u4fe1\u606f \u76ee\u6807\u7c7b\u522b： [\u20180\u2019, \u20181\u2019]\u4e2d\u6587\u7c7b\u522b：[\u2018\u91ce\u732a\u2019, \u2018\u91ce\u732a\u2019]\u8bad\u7ec3\u96c6：2014 \u5f20\u9a8c\u8bc1\u96c6：582 \u5f20\u6d4b\u8bd5\u96c6：289 \u5f20\u603b\u8ba1：2885 \u5f20 📄 data.yaml \u914d\u7f6e\u4fe1\u606f \u8be5\u6570\u636e\u96c6\u63d0\u4f9b\u4e86data.yaml\u6587\u4ef6，\u5185\u5bb9\u5982\u4e0b&#","og_url":"https:\/\/www.wsisp.com\/helps\/103088.html","og_site_name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","article_published_time":"2026-09-09T13:49:04+00:00","og_image":[{"url":"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260909134902-6aa163ce81dae.jpg"}],"author":"admin","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"admin","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"5 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.wsisp.com\/helps\/103088.html","url":"https:\/\/www.wsisp.com\/helps\/103088.html","name":"YOLO\u91ce\u5916\u6797\u5730\u91ce\u732a\u76ee\u6807\u68c0\u6d4b\u6570\u636e\u96c6-2885\u5f20 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","isPartOf":{"@id":"https:\/\/www.wsisp.com\/helps\/#website"},"datePublished":"2026-09-09T13:49:04+00:00","dateModified":"2026-09-09T13:49:04+00:00","author":{"@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41"},"breadcrumb":{"@id":"https:\/\/www.wsisp.com\/helps\/103088.html#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.wsisp.com\/helps\/103088.html"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.wsisp.com\/helps\/103088.html#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/www.wsisp.com\/helps"},{"@type":"ListItem","position":2,"name":"YOLO\u91ce\u5916\u6797\u5730\u91ce\u732a\u76ee\u6807\u68c0\u6d4b\u6570\u636e\u96c6-2885\u5f20"}]},{"@type":"WebSite","@id":"https:\/\/www.wsisp.com\/helps\/#website","url":"https:\/\/www.wsisp.com\/helps\/","name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","description":"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41","name":"admin","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/","url":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","contentUrl":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","caption":"admin"},"sameAs":["http:\/\/wp.wsisp.com"],"url":"https:\/\/www.wsisp.com\/helps\/author\/admin"}]}},"_links":{"self":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/103088","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/comments?post=103088"}],"version-history":[{"count":0,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/103088\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media\/103085"}],"wp:attachment":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media?parent=103088"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/categories?post=103088"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/tags?post=103088"},{"taxonomy":"topic","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/topic?post=103088"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}