YOLOv8 ByteTrack 车辆跟踪计数实战跨线统计完整流程这篇教程根据我复现车辆跟踪计数流程时整理重点演示视频准备、YOLOv8 检测、ByteTrack 跟踪、跨线计数和结果视频导出。本文整理自我的学习和项目复现过程尽量按实操顺序保留 notebook 的关键步骤同时把数据集获取方式调整为适合中文教程发布的写法。本文会重点跑通以下流程准备车辆计数视频安装 YOLOv8 和 ByteTrack封装跟踪器参数与匹配逻辑筛选车辆类别并运行检测按跨线方向统计车辆数量如果你正在系统学习目标检测、实例分割、OCR、多目标跟踪或视觉大模型建议收藏本文配套 notebook、示例图片和运行环境说明后续会继续整理。如果环境配置卡住可以在评论区说明具体报错。 文章目录YOLOv8 ByteTrack 车辆跟踪计数实战跨线统计完整流程⚙️ 检查环境 准备车辆视频 安装 YOLOv8 安装 ByteTrack 封装跟踪参数 安装 Supervision 加载检测模型️ 筛选车辆类别️ 单帧检测预览 设置计数线 生成计数视频 小结 同系列教程汇总⚙️ 检查环境确认 GPU 和工作目录。!nvidia-smiimportos HOMEos.getcwd()print(HOME) 准备车辆视频下载车辆计数示例视频也可以替换为自己的视频。%cd{HOME}!wget--load-cookies/tmp/cookies.txthttps://docs.google.com/uc?exportdownloadconfirm$(wget --quiet --save-cookies /tmp/cookies.txt --keep-session-cookies --no-check-certificate https://docs.google.com/uc?exportdownloadid1pz68D1Gsx80MoPg-_q-IbEdESEmyVLm- -O- | sed -rn s/.*confirm([0-9A-Za-z_]).*/\1\n/p)id1pz68D1Gsx80MoPg-_q-IbEdESEmyVLm--O vehicle-counting.mp4rm-rf/tmp/cookies.txtSOURCE_VIDEO_PATHf{HOME}/vehicle-counting.mp4 安装 YOLOv8安装 Ultralytics 并关闭匿名同步。# Pip install method (recommended)!pip installultralytics8.3.40fromIPythonimportdisplay display.clear_output()# 关闭 Ultralytics 匿名同步!yolo settings syncFalseimportultralytics ultralytics.checks() 安装 ByteTrack安装 ByteTrack 及相关依赖。%cd{HOME}!git clone https://github.com/ifzhang/ByteTrack.git%cd{HOME}/ByteTrack# 兼容旧版依赖!sed-is/onnx1.8.1/onnx1.9.0/grequirements.txt !pip3 install-q-r requirements.txt !python3 setup.py-q develop !pip install-q cython_bbox !pip install-q onemetric# 兼容旧版依赖!pip install-q loguru lap thopfromIPythonimportdisplay display.clear_output()importsys sys.path.append(f{HOME}/ByteTrack)importyoloxprint(yolox.__version__:,yolox.__version__) 封装跟踪参数定义 ByteTrack 参数和检测框匹配逻辑。fromyolox.tracker.byte_trackerimportBYTETracker,STrackfromonemetric.cv.utils.iouimportbox_iou_batchfromdataclassesimportdataclassdataclass(frozenTrue)classBYTETrackerArgs:track_thresh:float0.25track_buffer:int30match_thresh:float0.8aspect_ratio_thresh:float3.0min_box_area:float1.0mot20:boolFalse 安装 Supervision安装视频读取、标注和跨线计数工具。!pip install supervision0.1.0fromIPythonimportdisplay display.clear_output()importsupervisionprint(supervision.__version__:,supervision.__version__)fromsupervision.draw.colorimportColorPalettefromsupervision.geometry.dataclassesimportPointfromsupervision.video.dataclassesimportVideoInfofromsupervision.video.sourceimportget_video_frames_generatorfromsupervision.video.sinkimportVideoSinkfromsupervision.notebook.utilsimportshow_frame_in_notebookfromsupervision.tools.detectionsimportDetections,BoxAnnotatorfromsupervision.tools.line_counterimportLineCounter,LineCounterAnnotatorfromtypingimportListimportnumpyasnp# converts Detections into format that can be consumed by match_detections_with_tracks functiondefdetections2boxes(detections:Detections)-np.ndarray:returnnp.hstack((detections.xyxy,detections.confidence[:,np.newaxis]))# converts List[STrack] into format that can be consumed by match_detections_with_tracks functiondeftracks2boxes(tracks:List[STrack])-np.ndarray:returnnp.array([track.tlbrfortrackintracks],dtypefloat)# matches our bounding boxes with predictionsdefmatch_detections_with_tracks(detections:Detections,tracks:List[STrack])-Detections:ifnotnp.any(detections.xyxy)orlen(tracks)0:returnnp.empty((0,))tracks_boxestracks2boxes(trackstracks)ioubox_iou_batch(tracks_boxes,detections.xyxy)track2detectionnp.argmax(iou,axis1)tracker_ids[None]*len(detections)fortracker_index,detection_indexinenumerate(track2detection):ifiou[tracker_index,detection_index]!0:tracker_ids[detection_index]tracks[tracker_index].track_idreturntracker_ids 加载检测模型加载 YOLOv8 车辆检测模型。# 参数设置MODELyolov8x.ptfromultralyticsimportYOLO modelYOLO(MODEL)model.fuse()️ 筛选车辆类别只保留 car、motorcycle、bus、truck 等类别。# 类别 ID 到类别名的映射CLASS_NAMES_DICTmodel.model.names# 关注车辆类别car、motorcycle、bus、truckCLASS_ID[2,3,5,7]️ 单帧检测预览在单帧上检查检测框和类别标签。# 创建视频帧生成器generatorget_video_frames_generator(SOURCE_VIDEO_PATH)# create instance of BoxAnnotatorbox_annotatorBoxAnnotator(colorColorPalette(),thickness4,text_thickness4,text_scale2)# acquire first video frameiteratoriter(generator)framenext(iterator)# model prediction on single frame and conversion to supervision Detectionsresultsmodel(frame)detectionsDetections(xyxyresults[0].boxes.xyxy.cpu().numpy(),confidenceresults[0].boxes.conf.cpu().numpy(),class_idresults[0].boxes.cls.cpu().numpy().astype(int))# format custom labelslabels[f{CLASS_NAMES_DICT[class_id]}{confidence:0.2f}for_,confidence,class_id,tracker_idindetections]# annotate and display frameframebox_annotator.annotate(frameframe,detectionsdetections,labelslabels)%matplotlib inline show_frame_in_notebook(frame,(16,16)) 设置计数线定义跨线统计位置和输出视频路径。# 参数设置LINE_STARTPoint(50,1500)LINE_ENDPoint(3840-50,1500)TARGET_VIDEO_PATHf{HOME}/vehicle-counting-result.mp4VideoInfo.from_video_path(SOURCE_VIDEO_PATH) 生成计数视频逐帧检测、跟踪并统计车辆通过数量。fromtqdm.notebookimporttqdm# create BYTETracker instancebyte_trackerBYTETracker(BYTETrackerArgs())# create VideoInfo instancevideo_infoVideoInfo.from_video_path(SOURCE_VIDEO_PATH)# 创建视频帧生成器generatorget_video_frames_generator(SOURCE_VIDEO_PATH)# create LineCounter instanceline_counterLineCounter(startLINE_START,endLINE_END)# create instance of BoxAnnotator and LineCounterAnnotatorbox_annotatorBoxAnnotator(colorColorPalette(),thickness4,text_thickness4,text_scale2)line_annotatorLineCounterAnnotator(thickness4,text_thickness4,text_scale2)# open target video filewithVideoSink(TARGET_VIDEO_PATH,video_info)assink:# loop over video framesforframeintqdm(generator,totalvideo_info.total_frames):# model prediction on single frame and conversion to supervision Detectionsresultsmodel(frame)detectionsDetections(xyxyresults[0].boxes.xyxy.cpu().numpy(),confidenceresults[0].boxes.conf.cpu().numpy(),class_idresults[0].boxes.cls.cpu().numpy().astype(int))# filtering out detections with unwanted classesmasknp.array([class_idinCLASS_IDforclass_idindetections.class_id],dtypebool)detections.filter(maskmask,inplaceTrue)# tracking detectionstracksbyte_tracker.update(output_resultsdetections2boxes(detectionsdetections),img_infoframe.shape,img_sizeframe.shape)tracker_idmatch_detections_with_tracks(detectionsdetections,trackstracks)detections.tracker_idnp.array(tracker_id)# filtering out detections without trackersmasknp.array([tracker_idisnotNonefortracker_idindetections.tracker_id],dtypebool)detections.filter(maskmask,inplaceTrue)# format custom labelslabels[f#{tracker_id}{CLASS_NAMES_DICT[class_id]}{confidence:0.2f}for_,confidence,class_id,tracker_idindetections]# updating line counterline_counter.update(detectionsdetections)# annotate and display frameframebox_annotator.annotate(frameframe,detectionsdetections,labelslabels)line_annotator.annotate(frameframe,line_counterline_counter)sink.write_frame(frame) 小结这篇教程完整整理了YOLOv8 车辆跟踪计数的核心复现流程。实际操作时建议先确认 GPU、依赖版本、数据集路径和模型权重路径再逐段运行 notebook。后续我会继续按源项目顺序整理同系列中的目标检测、实例分割、OCR、多目标跟踪和视觉大模型教程。 同系列教程汇总Google Gemini 3.5 Flash 零样本目标检测教程从提示词到可视化结果GLM-OCR 文档识别实战教程从验证码、公式到车牌 OCRRF-DETR ByteTrack 多目标跟踪实战教程从命令行到 Python 视频轨迹可视化SAM 3 图像分割实战教程文本、框和点提示的多种分割方式YOLOv8 ByteTrack 车辆跟踪计数实战跨线统计完整流程