Automated defect detection
Catch defects the moment they appear. Across recent peer-reviewed studies, YOLO-based inspection reaches 99%+ accuracy and cuts manual inspection costs by up to 94.5%, on steel, PCBs, fabric, solar panels, and welds. Train it on your own defects and deploy it on the line.
深受全球领先组织信赖
How Ultralytics YOLO tackles defect detection
One model, every defect type
Ultralytics YOLO detects, localizes, and classifies defects in a single pass, and researchers have pushed it past 98% mAP on benchmarks from PCBs to steel to wafers. That means fewer escaped defects, less scrap, and faster time-to-market, on the hardware you already run.
- Subtle defects, caught: 98.8% mAP on tiny PCB defects, 98.2% on wafer surfaces (peer-reviewed).
- Line-speed inference: YOLO26 runs a frame in ~1.7 ms; weld studies report 2.3 ms per part.
- Trainable on your defects: fine-tune on your own images in hours; published models hit 99%+ accuracy.
- Deploy anywhere: export to 19 formats — TensorRT, ONNX, OpenVINO, CoreML — for edge or cloud.
体验 YOLO26 推理
拖放一张图像以查看实时目标检测
Deploy on the line or in the cloud
Train once, then export to 19 formats and run wherever inspection happens, from a camera-side edge box to a GPU server.
At the edge (line-side)
- NVIDIA Jetson via TensorRT: real-time inference with no internet connection.
- Intel CPUs via OpenVINO: deploy on the industrial PCs you already run.
- Hailo, Raspberry Pi & ARM: low-power, camera-side inspection stations.
- Fast enough for the line: published weld inspection runs at 2.3 ms per part.
In the cloud
- GPU autoscaling for high-volume, multi-camera inspection.
- Retrain centrally and push new defect classes to every line at once.
- Export to ONNX, TensorRT & PyTorch for any inference stack.
利用视觉 AI 变革行业
从工厂车间到手术室，Ultralytics 将视觉数据转化为实时决策。
SOHGA 借助 Ultralytics YOLO 将停车场监控时间缩短了 30%
Scaleout 借助 Ultralytics YOLO 将模型更新时间从数周缩短至数小时
RapiD Engineering 使用 Ultralytics YOLO 将海鲜质量控制的部署速度提高了 1 周
Project Ocean Oasis 利用 Ultralytics YOLO 推进珊瑚礁保护
Volley 利用 Ultralytics YOLO 为超过 250 个场上 AI 教练提供支持
WG Tech Solutions 通过 Ultralytics YOLO 和 Axelera 的 AI 加速器将安全违规行为减少了 28%
Stride 使用 Ultralytics YOLO 提供 1 分钟的马匹步态分析
Pixelabs 通过 Ultralytics YOLO 驱动的自动化实现了 95% 的召回率
SiteAssist 通过 Ultralytics YOLO 处理超过 77 万张图像来改善现场安全
Chef Robotics 使用 Ultralytics YOLO 将食品损耗降低了 67%
Cali Intelligence 使用 Ultralytics YOLO 将结账排队时间缩短了 43%
MarineSitu 使用 Ultralytics YOLO 在水下监测中实现 96%+ 的正常运行时间
Theia Scientific 使用 Ultralytics YOLO 将显微镜分析速度提高了 43 倍
eSmart Systems 使用 Ultralytics YOLO 将电力巡检时间缩短了一半
Axelera AI 使用 Ultralytics YOLO 实现 34 FPS 的边缘 AI 推理
STMicroelectronics 在 MCU 上运行 Ultralytics YOLO，每次推理仅需 9.4 mJ
Specialvideo 使用 Ultralytics YOLO 实现了 99% 的食品检测准确率
Vivity AI 通过 Ultralytics YOLO 每年在工业运营中节省超过 500 万美元
Videologic Analytics 通过 Ultralytics YOLO 扩展至 1 万个 AI 摄像头许可证
Prezent 使用 Ultralytics YOLO 将幻灯片检测准确率提高了 34%
ALYCE 通过 Ultralytics YOLO 将交通 AI 推理速度提升了 20%
Kiwitron 使用 Ultralytics YOLO 在 30 米外检测工业危险
SOHGA 借助 Ultralytics YOLO 将停车场监控时间缩短了 30%
Scaleout 借助 Ultralytics YOLO 将模型更新时间从数周缩短至数小时
RapiD Engineering 使用 Ultralytics YOLO 将海鲜质量控制的部署速度提高了 1 周
Project Ocean Oasis 利用 Ultralytics YOLO 推进珊瑚礁保护
Volley 利用 Ultralytics YOLO 为超过 250 个场上 AI 教练提供支持
WG Tech Solutions 通过 Ultralytics YOLO 和 Axelera 的 AI 加速器将安全违规行为减少了 28%
Stride 使用 Ultralytics YOLO 提供 1 分钟的马匹步态分析
Pixelabs 通过 Ultralytics YOLO 驱动的自动化实现了 95% 的召回率
SiteAssist 通过 Ultralytics YOLO 处理超过 77 万张图像来改善现场安全
Chef Robotics 使用 Ultralytics YOLO 将食品损耗降低了 67%
Cali Intelligence 使用 Ultralytics YOLO 将结账排队时间缩短了 43%
MarineSitu 使用 Ultralytics YOLO 在水下监测中实现 96%+ 的正常运行时间
Theia Scientific 使用 Ultralytics YOLO 将显微镜分析速度提高了 43 倍
eSmart Systems 使用 Ultralytics YOLO 将电力巡检时间缩短了一半
Axelera AI 使用 Ultralytics YOLO 实现 34 FPS 的边缘 AI 推理
STMicroelectronics 在 MCU 上运行 Ultralytics YOLO，每次推理仅需 9.4 mJ
Specialvideo 使用 Ultralytics YOLO 实现了 99% 的食品检测准确率
Vivity AI 通过 Ultralytics YOLO 每年在工业运营中节省超过 500 万美元
Videologic Analytics 通过 Ultralytics YOLO 扩展至 1 万个 AI 摄像头许可证
Prezent 使用 Ultralytics YOLO 将幻灯片检测准确率提高了 34%
ALYCE 通过 Ultralytics YOLO 将交通 AI 推理速度提升了 20%
Kiwitron 使用 Ultralytics YOLO 在 30 米外检测工业危险
SOHGA 借助 Ultralytics YOLO 将停车场监控时间缩短了 30%
Scaleout 借助 Ultralytics YOLO 将模型更新时间从数周缩短至数小时
RapiD Engineering 使用 Ultralytics YOLO 将海鲜质量控制的部署速度提高了 1 周
Project Ocean Oasis 利用 Ultralytics YOLO 推进珊瑚礁保护
Volley 利用 Ultralytics YOLO 为超过 250 个场上 AI 教练提供支持
WG Tech Solutions 通过 Ultralytics YOLO 和 Axelera 的 AI 加速器将安全违规行为减少了 28%
Stride 使用 Ultralytics YOLO 提供 1 分钟的马匹步态分析
Pixelabs 通过 Ultralytics YOLO 驱动的自动化实现了 95% 的召回率
SiteAssist 通过 Ultralytics YOLO 处理超过 77 万张图像来改善现场安全
Chef Robotics 使用 Ultralytics YOLO 将食品损耗降低了 67%
Cali Intelligence 使用 Ultralytics YOLO 将结账排队时间缩短了 43%
MarineSitu 使用 Ultralytics YOLO 在水下监测中实现 96%+ 的正常运行时间
Theia Scientific 使用 Ultralytics YOLO 将显微镜分析速度提高了 43 倍
eSmart Systems 使用 Ultralytics YOLO 将电力巡检时间缩短了一半
Axelera AI 使用 Ultralytics YOLO 实现 34 FPS 的边缘 AI 推理
STMicroelectronics 在 MCU 上运行 Ultralytics YOLO，每次推理仅需 9.4 mJ
Specialvideo 使用 Ultralytics YOLO 实现了 99% 的食品检测准确率
Vivity AI 通过 Ultralytics YOLO 每年在工业运营中节省超过 500 万美元
Videologic Analytics 通过 Ultralytics YOLO 扩展至 1 万个 AI 摄像头许可证
Prezent 使用 Ultralytics YOLO 将幻灯片检测准确率提高了 34%
ALYCE 通过 Ultralytics YOLO 将交通 AI 推理速度提升了 20%
Kiwitron 使用 Ultralytics YOLO 在 30 米外检测工业危险
常见问题解答
Automated defect detection uses cameras and AI to inspect products and surfaces for flaws, scratches, cracks, dents, missing parts, and contamination, in real time, with no manual checking. Ultralytics YOLO models detect, localize, and classify each defect in a single pass as parts move down the line.
It depends on the data, but recent peer-reviewed studies report 84-99% accuracy with YOLO models tuned to a specific task, for example 99.4% on fabric, 98.8% mAP on PCBs, and 98.2% on semiconductor wafers. The Ultralytics Platform supports active learning, so accuracy keeps climbing as new defect types appear.
Published 2024-2026 studies apply Ultralytics YOLO to steel and metal surfaces, printed circuit boards, semiconductor wafers, welds, textiles, solar panels, aluminum castings, ceramics, batteries, and pharmaceutical packaging, anywhere visual quality control matters.
Yes. Rule-based machine vision struggles with variable, subtle, or previously-unseen defects. Ultralytics YOLO26 learns from your own examples to recognize scratches, hairline cracks, white spots, and missing components across changing lighting and part variation, reaching 98%+ mAP on several published benchmarks.
Fast enough for the line. YOLO26 processes a frame in about 1.7 ms on a modern GPU; published inspection systems report 186 frames per second on fabric and 2.3 ms per weld.
Yes. Most inspection deployments run on-premises because production data is sensitive and factory networks are unreliable. Ultralytics YOLO26 exports to 19 formats, including TensorRT for NVIDIA Jetson, OpenVINO for Intel CPUs, and Hailo accelerators, so models run line-side with no internet connection required.
Less than you might expect. Published defect-detection models train on a few thousand labeled images, for example the bridge-crack study used a 4,029-image dataset. The Ultralytics Platform handles annotation and training, and pretrained YOLO weights mean you fine-tune rather than start from scratch.
Collect images of your good and defective parts, annotate them, and train, all in one place on the Ultralytics Platform. For production deployment, enterprise licensing covers commercial use and keeps your code and data private.