SH17
收藏资源简介:
SH17数据集由温莎大学的机械、汽车和材料工程系创建,包含8,099张标注图像,涵盖17类个人防护装备(PPE),如安全帽、安全眼镜等。数据集从多样化的工业环境中收集,旨在通过对象检测和卷积神经网络技术,提高制造业中的人身安全。创建过程中,数据通过Pexels网站收集并由专业人员进行标注,确保了数据的质量和多样性。该数据集主要用于训练和验证对象检测模型,以解决工业环境中的个人防护装备合规性问题。
The SH17 dataset was developed by the Department of Mechanical, Automotive, and Materials Engineering at the University of Windsor. It contains 8,099 annotated images spanning 17 classes of personal protective equipment (PPE), including hard hats, safety glasses, and other related protective gear. The dataset was gathered from diverse industrial environments, with the core objective of enhancing workplace safety in manufacturing through object detection and convolutional neural network technologies. During its development, the dataset's images were collected from the Pexels website and annotated by professional personnel, ensuring the dataset's quality and diversity. This dataset is primarily used for training and validating object detection models to address PPE compliance issues in industrial settings.
SH17: A Dataset for Human Safety and Personal Protective Equipment Detection in Manufacturing Industry
数据集概述
摘要
工作场所事故对人类安全构成重大风险,特别是在建筑和制造业等行业,个人防护装备(PPE)合规性的必要性日益突出。本研究基于目标检测(OD)和卷积神经网络(CNN)开发了非侵入式技术,用于检测和验证各种PPE(如头盔、安全眼镜、口罩和防护服)的正确使用。本研究提出了SH17数据集,包含8,099张标注图像,包含75,994个17类别的实例,收集自多样化的工业环境,用于训练和验证OD模型。我们训练了最先进的OD模型进行基准测试,初步结果显示YOLOv9-e模型变体在PPE检测中超过70.9%的准确率。模型在跨域数据集上的验证性能表明,集成这些技术可以显著提高安全管理系统,为努力满足人类安全法规和保护其劳动力的行业提供可扩展和高效的解决方案。
数据集详情
- 图像数量: 8,099张标注图像
- 实例数量: 75,994个对象实例
- 类别数量: 17类PPE物品和身体部位
关键特性
- 从全球多样化的工业环境中收集
- 高质量图像(最大分辨率8192x5462,最小1920x1002)
- 平均每张图像包含9.38个实例
- 包括小物体如耳朵和耳罩(39,764个标注 < 1%图像面积,59,025个标注 < 5%面积)
类别
- Person
- Head
- Face
- Glasses
- Face-mask-medical
- Face-guard
- Ear
- Earmuffs
- Hands
- Gloves
- Foot
- Shoes
- Safety-vest
- Tools
- Helmet
- Medical-suit
- Safety-suit
模型
我们使用ultralytics==8.0.38仓库训练了所有YOLO模型。我们提供了用于评估和基准测试的YOLO v8、v9和v10对象检测模型的训练权重。
YOLO v8, v9, v10 结果
| 模型 | 参数 | 图像数量 | 实例数量 | P | R | mAP50 | mAP50-95 |
|---|---|---|---|---|---|---|---|
| Yolo-8-n | 3.2 | 1620 | 15358 | 67.5 | 53.6 | 58.0 | 36.6 |
| Yolo-8-s | 11.2 | 1620 | 15358 | 81.5 | 55.7 | 63.7 | 41.7 |
| Yolo-8-m | 25.9 | 1620 | 15358 | 77.1 | 60.5 | 66.6 | 45.7 |
| Yolo-8-l | 43.7 | 1620 | 15358 | 76.7 | 62.9 | 68.0 | 47.0 |
| Yolo-8-x | 68.2 | 1620 | 15358 | 77.1 | 63.1 | 69.3 | 47.2 |
| Yolo-9-t | 2.0 | 1620 | 15358 | 75.0 | 52.6 | 58.5 | 37.5 |
| Yolo-9-s | 7.2 | 1620 | 15358 | 73.6 | 60.2 | 65.3 | 42.9 |
| Yolo-9-m | 20.1 | 1620 | 15358 | 77.4 | 62.0 | 68.6 | 46.5 |
| Yolo-9-c | 25.5 | 1620 | 15358 | 79.6 | 60.8 | 67.7 | 46.5 |
| Yolo-9-e | 58.1 | 1620 | 15358 | 81.0 | 65.0 | 70.9 | 48.7 |
| Yolo-10-n | 2.3 | 1620 | 15358 | 66.8 | 53.2 | 57.2 | 35.9 |
| Yolo-10-s | 7.2 | 1620 | 15358 | 75.8 | 57.0 | 62.7 | 40.9 |
| Yolo-10-m | 15.4 | 1620 | 15358 | 71.4 | 61.4 | 65.7 | 43.8 |
| Yolo-10-b | 19.1 | 1620 | 15358 | 77.7 | 59.1 | 65.8 | 45.1 |
| Yolo-10-l | 24.4 | 1620 | 15358 | 76.0 | 61.8 | 67.4 | 46.0 |
| Yolo-10-x | 29.5 | 1620 | 15358 | 76.8 | 62.8 | 67.8 | 46.7 |
使用方法
python from ultralytics import YOLO
model = YOLO(r"path oweight.pt") # 提供训练模型的路径 results = model(pathofimage) # 运行推理
results.show() # 查看结果
更多使用细节请参考ultralytics文档。
评估
我们提供了评估代码,以在SH17验证集上重现我们的基准测试结果。
python from ultralytics import YOLO
model = YOLO(r"path oweight.pt") model.val(data="sh17.yaml", batch = 1, imgsz = 640, device="cuda:0")
引用
如果您在研究中使用此数据集或代码,请引用我们的论文:
latex @article{ahmad_2024_sh17, title={SH17: A Dataset for Human Safety and Personal Protective Equipment Detection in Manufacturing Industry}, author={Ahmad, Hafiz Mughees and Rahimi, Afshin}, journal={Arxiv}, year={2024} }
许可证
SH17数据集在CC BY-NC-SA 4.0许可证下发布。

- 1SH17: A Dataset for Human Safety and Personal Protective Equipment Detection in Manufacturing Industry机械、汽车和材料工程系,温莎大学 · 2024年



