{ "source_file": "0930", "source_platform": "技术平台", "author": "Ultralytics", "title_listed": "Ultralytics YOLOv8 文档", "keywords": "YOLOv8,目标检测,分割", "domain": "图像处理", "content_type": "官方文档", "is_paid_flag": "否", "url": "https://docs.ultralytics.com/models/yolov8/", "publish_date": "持续更新", "corpus_stage": "图像处理", "notes": "YOLOv8 官方文档", "id": "faf874dd5204", "collected_at": "2026-10-02T18:25:17", "title": "Explore Ultralytics YOLOv8", "body_markdown": "Explore Ultralytics YOLOv8\n#\nYOLOv8\nultralytics\npublic\n8\nReleased January 2023. Proven and widely adopted with extensive community support. Introduced the unified Python API and CLI that became the Ultralytics standard.\n14\n25\nmodels\n1.4 GB\nMay 21, 2026\nyolov8n-obb\nyolov8x-obb\nyolov8l-obb\n+\n22\nOverview\n#\nYOLOv8 was released by Ultralytics on January 10, 2023, offering cutting-edge performance in terms of accuracy and speed. Building upon the advancements of previous YOLO versions, YOLOv8 introduced new features and optimizations that make it an ideal choice for various\nobject detection\ntasks in a wide range of applications.\nWatch:\nUltralytics YOLOv8 Model Overview\nTry on Ultralytics Platform\nExplore and run YOLOv8 models directly on\nUltralytics Platform\n.\nKey Features of YOLOv8\n#\nAdvanced Backbone and Neck Architectures:\nYOLOv8 employs state-of-the-art backbone and neck architectures, resulting in improved\nfeature extraction\nand\nobject detection\nperformance.\nAnchor-free Split Ultralytics Head:\nYOLOv8 adopts an anchor-free split Ultralytics head, which contributes to better accuracy and a more efficient detection process compared to anchor-based approaches.\nOptimized Accuracy-Speed Tradeoff:\nWith a focus on maintaining an optimal balance between accuracy and speed, YOLOv8 is suitable for real-time object detection tasks in diverse application areas.\nVariety of Pretrained Models:\nYOLOv8 offers a range of pretrained models to cater to various tasks and performance requirements, making it easier to find the right model for your specific use case.\nSupported Tasks and Modes\n#\nThe YOLOv8 series offers a diverse range of models, each specialized for specific tasks in computer vision. These models are designed to cater to various requirements, from object detection to more complex tasks like\ninstance segmentation\n, classification, pose/keypoints detection, and oriented object detection.\nEach variant of the YOLOv8 series is optimized for its respective task, ensuring high performance and accuracy. Additionally, these models are compatible with various operational modes including\nInference\n,\nValidation\n,\nTraining\n, and\nExport\n, facilitating their use in different stages of deployment and development.\nModel\nFilenames\nTask\nTraining\nValidation\nInference\nExport\nYOLOv8\nyolov8n.pt\nyolov8s.pt\nyolov8m.pt\nyolov8l.pt\nyolov8x.pt\nDetection\n✅\n✅\n✅\n✅\nYOLOv8-seg\nyolov8n-seg.pt\nyolov8s-seg.pt\nyolov8m-seg.pt\nyolov8l-seg.pt\nyolov8x-seg.pt\nInstance Segmentation\n✅\n✅\n✅\n✅\nYOLOv8-cls\nyolov8n-cls.pt\nyolov8s-cls.pt\nyolov8m-cls.pt\nyolov8l-cls.pt\nyolov8x-cls.pt\nClassification\n✅\n✅\n✅\n✅\nYOLOv8-pose\nyolov8n-pose.pt\nyolov8s-pose.pt\nyolov8m-pose.pt\nyolov8l-pose.pt\nyolov8x-pose.pt\nyolov8x-pose-p6.pt\nPose/Keypoints\n✅\n✅\n✅\n✅\nYOLOv8-obb\nyolov8n-obb.pt\nyolov8s-obb.pt\nyolov8m-obb.pt\nyolov8l-obb.pt\nyolov8x-obb.pt\nOriented Detection\n✅\n✅\n✅\n✅\nThis table provides an overview of the YOLOv8 model variants, highlighting their applicability in specific tasks and their compatibility with various operational modes such as Inference, Validation, Training, and Export. It showcases the versatility and robustness of the YOLOv8 series, making them suitable for a variety of applications in\ncomputer vision\n.\nPerformance Metrics\n#\nPerformance\nDetection (COCO)\nDetection (Open Images V7)\nSegmentation (COCO)\nClassification (ImageNet)\nPose (COCO)\nOBB (DOTAv1)\nSee\nDetection Docs\nfor usage examples with these models trained on\nCOCO\n, which include 80 pretrained classes.\nModel\nsize\n(pixels)\nmAP\nval\n50-95\nSpeed\nCPU ONNX\n(ms)\nSpeed\nA100 TensorRT\n(ms)\nparams\n(M)\nFLOPs\n(B)\nYOLOv8n\n640\n37.3\n80.4\n0.99\n3.2\n8.7\nYOLOv8s\n640\n44.9\n128.4\n1.20\n11.2\n28.6\nYOLOv8m\n640\n50.2\n234.7\n1.83\n25.9\n78.9\nYOLOv8l\n640\n52.9\n375.2\n2.39\n43.7\n165.1\nYOLOv8x\n640\n53.9\n479.1\n3.53\n68.2\n257.8\nYOLOv8 Usage Examples\n#\nThis example provides simple YOLOv8 training and inference examples. For full documentation on these and other\nmodes\nsee the\nPredict\n,\nTrain\n,\nVal\nand\nExport\ndocs pages.\nNote the below example is for YOLOv8\nDetect\nmodels for object detection. For additional supported tasks see the\nSegment\n,\nClassify\n,\nPose\n, and\nOBB\ndocs.\nExample\nPython\nCLI\nPyTorch\npretrained\n*.pt\nmodels as well as configuration\n*.yaml\nfiles can be passed to the\nYOLO()\nclass to create a model instance in Python:\nfrom ultralytics import YOLO\n\n# Load a COCO-pretrained YOLOv8n model\nmodel = YOLO(\"yolov8n.pt\")\n\n# Display model information (optional)\nmodel.info()\n\n# Train the model on the COCO8 example dataset for 100 epochs\nresults = model.train(data=\"coco8.yaml\", epochs=100, imgsz=640)\n\n# Run inference with the YOLOv8n model on the 'bus.jpg' image\nresults = model(\"path/to/bus.jpg\")\nCitations and Acknowledgments\n#\nUltralytics YOLOv8 Publication\nUltralytics has not published a formal research paper for YOLOv8 due to the rapidly evolving nature of the models. We focus on advancing the technology and making it easier to use, rather than producing static documentation. For the most up-to-date information on YOLO architecture, features, and usage, please refer to our\nGitHub repository\nand\ndocumentation\n.\nIf you use the YOLOv8 model or any other software from this repository in your work, please cite it using the following format:\nQuote\nBibTeX\n@software{yolov8_ultralytics,\n author = {Glenn Jocher and Ayush Chaurasia and Jing Qiu},\n title = {Ultralytics YOLOv8},\n version = {8.0.0},\n year = {2023},\n url = {https://github.com/ultralytics/ultralytics},\n orcid = {0000-0001-5950-6979, 0000-0002-7603-6750, 0000-0003-3783-7069},\n license = {AGPL-3.0}\n}\nPlease note that the DOI is pending and will be added to the citation once it is available. YOLOv8 models are provided under\nAGPL-3.0\nand\nEnterprise\nlicenses.\nFAQ\n#\nWhat is YOLOv8 and how does it differ from previous YOLO versions?\nYOLOv8 is designed to improve real-time object detection performance with advanced features. Unlike earlier versions, YOLOv8 incorporates an\nanchor-free split Ultralytics head\n, state-of-the-art\nbackbone\nand neck architectures, and offers optimized\naccuracy\n-speed tradeoff, making it ideal for diverse applications. For more details, check the\nOverview\nand\nKey Features\nsections.\nHow can I use YOLOv8 for different computer vision tasks?\nYOLOv8 supports a wide range of computer vision tasks, including object detection, instance segmentation, classification, pose/keypoints detection, and oriented object detection. Each model variant is optimized for its specific task and compatible with various operational modes like\nInference\n,\nValidation\n,\nTraining\n, and\nExport\n. Refer to the\nSupported Tasks and Modes\nsection for more information.\nWhat are the performance metrics for YOLOv8 models?\nYOLOv8 models achieve state-of-the-art performance across various benchmarking datasets. For instance, the YOLOv8n model achieves a mAP (mean Average Precision) of 37.3 on the COCO dataset and a speed of 0.99 ms on A100 TensorRT. Detailed performance metrics for each model variant across different tasks and datasets can be found in the\nPerformance Metrics\nsection.\nHow do I train a YOLOv8 model?\nTraining a YOLOv8 model can be done using either Python or CLI. Below are examples for training a model using a COCO-pretrained YOLOv8 model on the COCO8 dataset for 100\nepochs\n:\nExample\nPython\nCLI\nfrom ultralytics import YOLO\n\n# Load a COCO-pretrained YOLOv8n model\nmodel = YOLO(\"yolov8n.pt\")\n\n# Train the model on the COCO8 example dataset for 100 epochs\nresults = model.train(data=\"coco8.yaml\", epochs=100, imgsz=640)\nFor further details, visit the\nTraining\ndocumentation.\nCan I benchmark YOLOv8 models for performance?\nYes, YOLOv8 models can be benchmarked for performance in terms of speed and accuracy across various export formats. You can use PyTorch, ONNX, TensorRT, and more for benchmarking. Below are example commands for benchmarking using Python and CLI:\nExample\nPython\nCLI\nfrom ultralytics.utils.benchmarks import benchmark\n\n# Benchmark on GPU\nbenchmark(model=\"yolov8n.pt\", data=\"coco8.yaml\", imgsz=640, device=0)\nFor additional information, check the\nPerformance Metrics\nsection.\nContributors\nGL\nglenn-jocher\n34\nLA\nLaughing-q\n4\nRI\nRizwanMunawar\n4\nRA\nraimbekovm\n3\nY-\nY-T-G\n3\nPD\npderrenger\n2\nON\nonuralpszr\n1\nLE\nLexBarou\n1\nMA\nMatthewNoyce\n1\nBU\nBurhan-Q\n1\nAY\nAyushExel\n1\nFC\nfcakyon\n1\nCreated\nNov 12, 2023\nUpdated\n2 weeks ago\nComments", "metadata": { "title": "Explore Ultralytics YOLOv8", "description": "Discover Ultralytics YOLOv8, an advancement in real-time object detection, optimizing performance with an array of pretrained models for diverse tasks.", "keywords": "YOLOv8,real-time object detection,YOLO series,Ultralytics,computer vision,advanced object detection,AI,machine learning,deep learning", "author": null, "og:title": "Explore Ultralytics YOLOv8 | Ultralytics Docs", "og:description": "Discover Ultralytics YOLOv8, an advancement in real-time object detection, optimizing performance with an array of pretrained models for diverse tasks.", "og:url": "https://docs.ultralytics.com/models/yolov8", "og:site_name": "Ultralytics Docs", "og:locale": "en_US", "og:image": "https://cdn.ul.run/i/7dd03f050503e5700d4cf85d6ea7ae3f.webp", "og:locale:alternate": "ar_SA", "og:type": "article", "twitter:card": "summary_large_image", "twitter:title": "Explore Ultralytics YOLOv8 | Ultralytics Docs", "twitter:description": "Discover Ultralytics YOLOv8, an advancement in real-time object detection, optimizing performance with an array of pretrained models for diverse tasks.", "twitter:image": "https://cdn.ul.run/i/7dd03f050503e5700d4cf85d6ea7ae3f.webp" }, "paywall": { "listed_paid": false, "detected": false, "marker": "", "public_only": false, "policy": "付费/会员内容只保留页面公开可见部分，不绕过付费墙" }, "crawl": { "success": true, "status_code": 200, "error": "" }, "code_blocks": [ { "file": "attachments/faf874dd5204/code/body-000.txt", "chars": 373, "language": "text" }, { "file": "attachments/faf874dd5204/code/body-001.txt", "chars": 313, "language": "text" }, { "file": "attachments/faf874dd5204/code/body-002.txt", "chars": 222, "language": "text" }, { "file": "attachments/faf874dd5204/code/body-003.txt", "chars": 140, "language": "text" } ], "images": [ { "status": "ok", "url": "https://cdn.ul.run/i/7dd03f050503e5700d4cf85d6ea7ae3f.avif", "bytes": 47037, "reused": true, "local_path": "attachments/faf874dd5204/images/000.img", "src": "https://cdn.ul.run/i/7dd03f050503e5700d4cf85d6ea7ae3f.avif", "alt": "YOLOv8 performance comparison charts", "downloaded": true } ], "audio": [], "videos": [ { "url": "https://www.youtube.com/watch?v=Na0HvJ4hkk0", "src": "https://www.youtube-nocookie.com/embed/Na0HvJ4hkk0", "status": "ok", "kind": "single_video", "local_path": "attachments/faf874dd5204/videos/repair-Na0HvJ4hkk0.mp4", "bytes": 20800359, "valid": true, "height": 1080, "width": 1920, "duration": "290.181000", "reused": true, "reused_from": "xinxing-zimeiti-corpus/attachments/c0a86fefcdd4/videos/Na0HvJ4hkk0.mp4" } ], "links": [ { "href": "https://github.com/ultralytics/ultralytics/blob/main/docs/en/models/yolov8.md", "label": "", "type": "a" }, { "href": "https://platform.ultralytics.com/ultralytics/yolov8?utm_source=docs&utm_medium=card", "label": "YOLOv8 ultralytics public 8 Released January 2023. Proven and widely adopted with extensive community support. Introduce", "type": "a" }, { "href": "https://www.ultralytics.com/blog/a-guide-to-deep-dive-into-object-detection-in-2025", "label": "object detection", "type": "a" }, { "href": "https://www.youtube-nocookie.com/embed/Na0HvJ4hkk0", "label": "", "type": "iframe" }, { "href": "https://platform.ultralytics.com/ultralytics/yolov8", "label": "Ultralytics Platform", "type": "a" }, { "href": "https://www.ultralytics.com/glossary/feature-extraction", "label": "feature extraction", "type": "a" }, { "href": "https://www.ultralytics.com/glossary/object-detection", "label": "object detection", "type": "a" }, { "href": "https://www.ultralytics.com/glossary/instance-segmentation", "label": "instance segmentation", "type": "a" }, { "href": "https://docs.ultralytics.com/modes/predict", "label": "Inference", "type": "a" }, { "href": "https://docs.ultralytics.com/modes/val", "label": "Validation", "type": "a" }, { "href": "https://docs.ultralytics.com/modes/train", "label": "Training", "type": "a" }, { "href": "https://docs.ultralytics.com/modes/export", "label": "Export", "type": "a" }, { "href": "https://docs.ultralytics.com/tasks/detect", "label": "Detection", "type": "a" }, { "href": "https://docs.ultralytics.com/tasks/segment", "label": "Instance Segmentation", "type": "a" }, { "href": "https://docs.ultralytics.com/tasks/classify", "label": "Classification", "type": "a" }, { "href": "https://docs.ultralytics.com/tasks/pose", "label": "Pose/Keypoints", "type": "a" }, { "href": "https://docs.ultralytics.com/tasks/obb", "label": "Oriented Detection", "type": "a" }, { "href": "https://www.ultralytics.com/glossary/computer-vision-cv", "label": "computer vision", "type": "a" }, { "href": "https://docs.ultralytics.com/datasets/detect/coco", "label": "COCO", "type": "a" }, { "href": "https://platform.ultralytics.com/ultralytics/yolov8/yolov8n", "label": "YOLOv8n", "type": "a" }, { "href": "https://platform.ultralytics.com/ultralytics/yolov8/yolov8s", "label": "YOLOv8s", "type": "a" }, { "href": "https://platform.ultralytics.com/ultralytics/yolov8/yolov8m", "label": "YOLOv8m", "type": "a" }, { "href": "https://platform.ultralytics.com/ultralytics/yolov8/yolov8l", "label": "YOLOv8l", "type": "a" }, { "href": "https://platform.ultralytics.com/ultralytics/yolov8/yolov8x", "label": "YOLOv8x", "type": "a" }, { "href": "https://docs.ultralytics.com/modes", "label": "modes", "type": "a" }, { "href": "https://www.ultralytics.com/glossary/pytorch", "label": "PyTorch", "type": "a" }, { "href": "https://github.com/ultralytics/ultralytics", "label": "GitHub repository", "type": "a" }, { "href": "https://docs.ultralytics.com/", "label": "documentation", "type": "a" }, { "href": "https://github.com/ultralytics/ultralytics/blob/main/LICENSE", "label": "AGPL-3.0", "type": "a" }, { "href": "https://www.ultralytics.com/license", "label": "Enterprise", "type": "a" }, { "href": "https://www.ultralytics.com/glossary/backbone", "label": "backbone", "type": "a" }, { "href": "https://www.ultralytics.com/glossary/accuracy", "label": "accuracy", "type": "a" }, { "href": "https://www.ultralytics.com/glossary/epoch", "label": "epochs", "type": "a" }, { "href": "https://github.com/glenn-jocher", "label": "GL glenn-jocher 34", "type": "a" }, { "href": "https://github.com/Laughing-q", "label": "LA Laughing-q 4", "type": "a" }, { "href": "https://github.com/RizwanMunawar", "label": "RI RizwanMunawar 4", "type": "a" }, { "href": "https://github.com/raimbekovm", "label": "RA raimbekovm 3", "type": "a" }, { "href": "https://github.com/Y-T-G", "label": "Y- Y-T-G 3", "type": "a" }, { "href": "https://github.com/pderrenger", "label": "PD pderrenger 2", "type": "a" }, { "href": "https://github.com/onuralpszr", "label": "ON onuralpszr 1", "type": "a" }, { "href": "https://github.com/LexBarou", "label": "LE LexBarou 1", "type": "a" }, { "href": "https://github.com/MatthewNoyce", "label": "MA MatthewNoyce 1", "type": "a" }, { "href": "https://github.com/Burhan-Q", "label": "BU Burhan-Q 1", "type": "a" }, { "href": "https://github.com/AyushExel", "label": "AY AyushExel 1", "type": "a" }, { "href": "https://github.com/fcakyon", "label": "FC fcakyon 1", "type": "a" } ], "repair_version": 2, "repaired_at": "2026-10-14T10:54:54Z", "source_url": "https://docs.ultralytics.com/models/yolov8/", "papers": [], "actual_video_files": [], "body_chars": 8276, "expected_body_images": 1, "discovery_status": "ok", "delivery_status": "ok", "missing_reason": "", "raw_html": "attachments/faf874dd5204/pages/source.html", "local_html": "attachments/faf874dd5204/pages/article.html", "body_base_path": "attachments/faf874dd5204/pages/", "body_markdown_path": "attachments/faf874dd5204/pages/article.md" }