当前位置：首页 > news >正文

工业表面缺陷检测开源数据集汇总

news 2025/6/30 15:15:45

开源数据集超级汇总：一文看尽工业表面缺陷检测 🚀

前言

在飞速发展的人工智能和机器视觉领域，高质量、公开的基准数据集是推动算法创新、技术落地和行业发展的核心基石。它们不仅为研究人员提供了验证新思想的“试金石”，也为开发者提供了训练和优化模型的宝贵资源。

本文旨在打造一份详尽的工业表面缺陷检测领域的“资源地图” ，全面汇总了由NEU机器视觉与机器人实验室在其系列文章（2025年6月16日至25日发布）中整理的公开数据集。内容涵盖了金属产品、非金属产品、通用异常检测以及路面缺陷检测四大专题。

希望通过这次系统性的梳理，为从事工业视觉、质量控制、自动化等领域的研究者和工程师们提供便利。

以下内容原始信息由“NEU机器视觉与机器人实验室”整理，本博客在原文基础上进行了综合、修订与完善。
我们对所有为构建和开放这些宝贵数据集的学者和机构，致以最崇高的敬意！ 🌸🌟

专题一：金属产品表面缺陷 🔩

本专题聚焦于各类金属制品的表面缺陷，是工业领域中应用最广泛的场景之一。

1. 带钢 (Strip Steel) 🏭

NEU-DET
- 适用任务: 🎯 目标检测
- 获取链接: https://pan.baidu.com/share/init?surl=nBbO-jWDm1_NHDQsc1dRkg (提取码: pmqx)
- 相关论文: An End-to-end Steel Surface Defect Detection Approach via Fusing Multiple Hierarchical Features
GC10-DET
- 适用任务: 🎯 目标检测
- 获取链接: https://www.kaggle.com/datasets/alex000kim/gc10det
NEU-Seg
- 适用任务: 🎨 语义分割
- 获取链接: https://github.com/DHW-Master/NEU_Seg
FSSD-12
- 适用任务: 🔬 小样本语义分割
- 获取链接: https://github.com/VDT-2048/CPANet
- 相关论文: Cross Position Aggregation Network for Few-shot Strip Steel Surface Defect Segmentation
Surface Defects-4i
- 适用任务: 🔬 小样本语义分割
- 获取链接: https://github.com/bbbbby-99/TGRNet-Surface-Defect-Segmentation
- 相关论文: Triplet-Graph Reasoning Network for Few-shot Metal Generic Surface Defect Segmentation
SEVERSTAL
- 适用任务: 🎨 语义分割
- 获取链接: https://www.kaggle.com/c/severstal-steel-defect-detection/data
- 相关论文: Severstal: Steel Defect Detection Dataset
NEU-CLS
- 适用任务: 🏷️ 分类
- 获取链接: https://pan.baidu.com/s/1l_RjTP7aTwr57ahcwelTpA
FSC / MSD-Cls
- 适用任务: 🔬 小样本分类
- 获取链接: https://github.com/successhaha/Gtnet
- 相关论文: Graph Embedding and Optimal Transport for Few-Shot Classification of Metal Surface Defect
FSC-20
- 适用任务: 🔬 小样本分类
- 获取链接: https://github.com/VDT-2048/FSC-20
- 相关论文: FaNet: Feature-aware Network for Few Shot Classification of Strip Steel Surface Defects
SD-saliency-900
- 适用任务: ✨ 显著性检测
- 获取链接: https://github.com/SongGuorong/MCITF/tree/master/SD-saliency-900
- 相关论文: EDRnet: Encoder–decoder residual network for salient object detection of strip steel surface defects
ESDis-SOD
- 适用任务: ✨ 显著性检测
- 获取链接: https://github.com/VDT-2048/A3Net
- 相关论文: Autocorrelation Aware Aggregation Network for Salient Object Detection of Strip Steel Surface Defects
Oil pollution defect
- 适用任务: ✨ 显著性检测
- 获取链接: https://pan.baidu.com/s/1_aU_Bfh7lcxpYW1no2MlUQ
- 相关论文: Surface defect detection method using saliency linear scanning morphology for silicon steel strip under oil pollution interference
Micro surface defect
- 适用任务: ✨ 显著性检测
- 获取链接: https://pan.baidu.com/s/1QM0AxlGjUlkHHyxwamIMmA
- 相关论文: Micro surface defect detection method for silicon steel strip based on saliency convex active contour model

2. 钢轨 (Steel Rail) 🚆

NRSD-MN
- 适用任务: ✨ 显著性检测 (RGB)
- 获取链接: https://github.com/zdfcvsn/Mcnet
- 相关论文: MCnet: Multiple Context Information Segmentation Network of No-service Rail Surface Defects
RSDDS-113
- 适用任务: 📷 联合检测 (RGB-D)
- 获取链接: https://github.com/neu-rail-rsdds/rsdds
- 相关论文: RSDDS: An Unsupervised Saliency Detection Dataset for Rail Surface Defects
NEU RSDDS-AUG
- 适用任务: 📷 联合检测 (RGB-D)
- 获取链接: https://github.com/VDT-2048/SAINet
- 相关论文: Collaborative Learning Attention Network Based on RGB Image and Depth Image for Surface Defect Inspection

3. 钢管 (Steel Pipe)

SSP2000
- 适用任务: ✨ 显著性检测
- 获取链接: https://github.com/VDT-2048/SRPCNet
- 相关论文: SR-PC-Net: A Two-Stage Network for Accurate Defect Detection of Seamless Steel Pipe Surface
CGFSDS-9
- 适用任务: 🔬 跨粒度小样本分割
- 获取链接: https://github.com/VDT-2048/MFANet
- 相关论文: MFANet: Multifeature Aggregation Network for Cross-granularity Few-shot Seamless Steel Tubes Surface Defect Segmentation

4. 其他金属制品

✈️ 航空发动机叶片 - ISAEB
- 适用任务: 🎨 分割
- 获取链接: https://github.com/Newbiejy/EGCIENet_In-service-blade-defect-detection
- 相关论文: An edge-guided defect segmentation network for in-service aerospace engine blades
✈️ 航空发动机叶片 - 涡轮叶片缺陷数据集
- 适用任务: 🎯 目标检测
- 获取链接: https://tianchi.aliyun.com/dataset/154228
🧲 磁瓦 - MTD
- 适用任务: ✨ 显著性检测
- 获取链接: https://github.com/Charmve/Surface-Defect-Detection/tree/master/Magnetic-Tile-Defect
- 相关论文: Surface defect saliency of magnetic tile
🏗️ 铝型材 - 天池铝型材表面缺陷
- 适用任务: 🎯 目标检测
- 获取链接: https://tianchi.aliyun.com/competition/entrance/231682/information
🔧 工业磨损 - BSData
- 适用任务: 🎨 实例分割
- 获取链接: https://github.com/2Obe/BSData
🔩 滚珠丝杠驱动器 - Ball Screw Drive Dataset
- 适用任务: 🏷️ 分类
- 获取链接: https://publikationen.bibliothek.kit.edu/10001338
⚙️ 齿轮 - GID
- 适用任务: 🎯 目标检测
- 获取链接: http://www.aiinnovation.com.cn/#/dataDetail?id=34
💡 光度立体视觉 - MSDD
- 适用任务: 🎯 目标检测
- 获取链接: https://www.scidb.cn/en/detail?dataSetId=3d739ddb4bdc439a9bf7ef550cae48d8
- 相关论文: A dataset for surface defect detection on complex structured parts based on photometric stereo
🔥 铸件 - CSDD
- 适用任务: 🎯 目标检测
- 获取链接: https://github.com/Kerio99/CSDD
- 相关论文: CSDD: A Benchmark Dataset for Casting Surface Defect Dete
🚗 汽车 - NCAT12-DET
- 适用任务: 🎯 目标检测
- 获取链接: https://github.com/Brym-Gyimah/NCAT12-DET
- 相关论文: NCAT12-DET: A New Benchmark Dataset for Surface Defect Detection and a Comparative Study

专题二：非金属产品表面缺陷 🧱

非金属材料如纺织品、陶瓷、塑料、木材等同样存在多样的缺陷形式，其检测技术在轻工业和高科技制造业中至关重要。

👕 织物 - 天池纺织品缺陷检测
- 适用任务: 🎯 目标检测
- 获取链接: https://tianchi.aliyun.com/dataset/94213
👕 织物 - TILDA
- 适用任务: 🏷️/🎯/⚠️ 分类/检测/异常
- 获取链接: https://lmb.informatik.uni-freiburg.de/resources/datasets/tilda.en.html
👕 织物 - AITEX FABRIC DB
- 适用任务: 🏷️/🎯/⚠️ 分类/检测/异常
- 获取链接: https://www.aitex.es/afid/
🧱 瓷砖 - 天池瓷砖瑕疵检测
- 适用任务: 🎯 目标检测
- 获取链接: https://tianchi.aliyun.com/dataset/110088
📱 手机屏幕 - PKU-Market-Phone
- 适用任务: 🎨 分割
- 获取链接: https://robotics.pkusz.edu.cn/resources/dataset/
🔌 PCB - PKU-Market-PCB
- 适用任务: 🎯/🏷️/📍 检测/分类/配准
- 获取链接: https://robotics.pkusz.edu.cn/resources/dataset/
🔌 PCB - FPCB-DET
- 适用任务: 🎯 目标检测
- 获取链接: https://github.com/Wesliee/decoupled-two-stage-framework
- 相关论文: A Decoupled Two-Stage Object Detection Framework
🔌 PCB - DsPCBSD+
- 适用任务: 🎯 目标检测
- 获取链接: https://figshare.com/articles/dataset/DsPCBSD_/24970329
🔌 PCB - DeepPCB
- 适用任务: 🎯/🏷️ 目标检测/分类
- 获取链接: https://github.com/Charmve/Surface-Defect-Detection/tree/master/DeepPCB
- 相关论文: A New PCB Defect Dataset and A Novel On-line Detection Method
⚡ 电力线绝缘子 - CPLID
- 适用任务: 🎯 目标检测
- 获取链接: https://github.com/InsulatorData/InsulatorDataSet
- 相关论文: Detection of Power Line Insulator Defects Using Aerial Images Analyzed With Convolutional Neural Networks
🪜 自动扶梯台阶 - 自动扶梯台阶缺陷
- 适用任务: 🎯 目标检测
- 获取链接: https://aistudio.baidu.com/aistudio/datasetdetail/44820
📦 工业产品 - 小样本工业缺陷检测
- 适用任务: 🔬 小样本分割
- 获取链接: https://www.kaggle.com/datasets/aryashah2k/few-shot-industrial-defect-detection
🪵 木材 - 木材表面缺陷
- 适用任务: 🎨 语义分割
- 获取链接: https://zenodo.org/records/4694695
☀️ 光伏板 - PV-Multi-Defect
- 适用任务: 🎯 目标检测
- 获取链接: https://github.com/CCNUZFW/PV-Multi-Defect
- 相关论文: GBH-YOLOv5: Ghost Convolution with BottleneckCSP and Tiny Target Prediction Head Incorporating YOLOv5 for PV Panel Defect Detection

专题三：异常检测 (Anomaly Detection) ⚠️

异常检测是工业质检的核心任务之一，旨在识别与正常样本有显著差异的缺陷。此类数据集通常包含大量正常样本和少量异常样本。

1. 单模态异常检测

NEU-RSDD-1, NEU-RSDD-2 (钢轨)
- 获取链接: https://github.com/neu-rail-rsdds/rail_surface_anomaly_detection.git
- 相关论文: An Adaptive Pyramid Graph and Variation Residual-Based Anomaly Detection Network for Rail Surface Defects
KolektorSDD (电子换向器)
- 获取链接: https://www.vicos.si/resources/kolektorsdd/
- 相关论文: Deep-learning-based computer vision system for surface-defect detection
KolektorSDD2 (电子换向器)
- 获取链接: https://www.vicos.si/resources/kolektorsdd2/
AeBAD (航空发动机叶片)
- 获取链接: https://github.com/zhangzilongc/MMR
- 相关论文: Industrial Anomaly Detection with Domain Shift: A Real-world Dataset and Masked Multi-scale Reconstruction
elpv-dataset (太阳能电池板)
- 获取链接: https://github.com/zae-bayern/elpv-dataset
- 相关论文: A Benchmark for Visual Identification of Defective Solar Cells in Electroluminescence Imagery
3CAD (3C产品)
- 获取链接: https://github.com/EnquanYang2022/3CAD
- 相关论文: 3CAD: A Large-scale Real-World 3C product Dataset for Unsupervised Anomaly Detection
CID (绝缘子)
- 获取链接: https://github.com/Light-ZhangTao/Insulator-Defect-Detection
- 相关论文: Catenary Insulator Defect Detection: A Dataset and an Unsupervised Baseline
MIAD (工业维护)
- 获取链接: https://miad-2022.github.io/
MPDD (金属零件)
- 获取链接: https://github.com/stepanje/MPDD
- 相关论文: Deep learning-based defect detection of metal parts: evaluating current methods in complex conditions
MVTec AD (工业通用)
- 获取链接: https://www.mvtec.com/company/research/datasets/mvtec-ad
- 相关论文: The MVTec Anomaly Detection Dataset: A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection
MVTec LOCO-AD (逻辑约束)
- 获取链接: https://www.mvtec.com/company/research/datasets/mvtec-loco
- 相关论文: BeyondDentsandScratches:Logical Constraints in Unsupervised AnomalyDetection andLocalization
VISION (工业通用)
- 获取链接: https://huggingface.co/datasets/VISION-Workshop/VISION-Datasets
- 相关论文: VISION Datasets: A Benchmark for Vision-based InduStrial InspectiON
Real-IAD (工业通用)
- 获取链接: https://realiad4ad.github.io/Real-IAD/
- 相关论文: Real-IAD: A Real-World Multi-View Dataset for Benchmarking Versatile Industrial Anomaly Detection
RAD (图像通用)
- 获取链接: https://github.com/hustCYQ/RAD-dataset
- 相关论文: RAD: A Comprehensive Dataset for Benchmarking the Robustness of Image Anomaly Detection
VisA (通用)
- 获取链接: https://github.com/amazon-science/spot-diff
- 相关论文: SPot-the-Difference Self-Supervised Pre-training for Anomaly Detection and Segmentation
IPAS, AS-Crack500, AS-GAPs (路面)
- 获取链接: https://github.com/Yaan-Wang/Pavement-Defect-Datasets.git
- 相关论文: Normal-knowledge-based pavement defect segmentation using relevance-aware and cross-reasoning mechanisms

2. 多模态异常检测 🧊

Eyecandies
- 获取链接: https://eyecan-ai.github.io/eyecandies
- 相关论文: The Eyecandies Dataset for Unsupervised Multimodal Anomaly Detection and Localization
MVTec 3D-AD
- 获取链接: https://www.mvtec.com/company/research/datasets/mvtec-3d-ad
- 相关论文: The MVTec3D-ADDataset for Unsupervised 3D Anomaly Detection and Localization
Real3D-AD
- 获取链接: https://github.com/M-3LAB/Real3D-AD
- 相关论文: Real3D-AD: A Dataset of Point Cloud Anomaly Detection
Real-IAD D3
- 获取链接: https://realiad4ad.github.io/Real-IAD_D3/
- 相关论文: Real-IAD D3: A Real-World 2D/Pseudo-3D/3D Dataset for Industrial Anomaly Detection
MulSen-AD
- 获取链接: https://zzzbbbzzz.github.io/MulSen_AD/index.html
- 相关论文: Multi-Sensor Object Anomaly Detection: Unifying Appearance, Geometry, and Internal Properties

专题四：路面缺陷检测 🛣️

路面缺陷检测对于交通安全、基础设施维护至关重要，是计算机视觉在土木工程领域的典型应用。

1. 分类 (Classification) ✅

GAPs v1 / v2
- 获取链接: https://www.tu-ilmenau.de/neurob/data-sets-code/gaps/
- 相关论文: How to Get Pavement Distress Detection Ready for Deep Learning? A Systematic Approach & Improving Visual Road Condition Assessment by Extensive Experiments on the Extended GAPs Dataset
LRRD
- 获取链接: https://www.designsafe-ci.org/data/browser/public/designsafe.storage.published/PRJ-2681
- 相关论文: Laser-scanned range image dataset from asphalt and concrete roadways for dcnn-based crack classification
CQU-BPMDD
- 获取链接: https://github.com/ychxff/CQU-BPMDD
- 相关论文: Deep Domain Adaptation for Pavement Crack Detection
CQU-BPDD
- 获取链接: https://dearcaat.github.io/CQU-BPDD
- 相关论文: An iteratively optimized patch label inference network for automatic pavement distress detection

2. 目标检测 (Object Detection) 🎯

RDD-2018/19/20/22
- 获取链接: https://github.com/sekilab/RoadDamageDetector.git
- 相关论文涵盖: Road Damage Detection and Classification Using Deep Neural Networks…, Generative adversarial network for road damage detection, etc.
PID
- 获取链接: https://github.com/Nan2020/PID-Pavement-Image-Dataset.git
- 相关论文: PID: A New Benchmark Dataset to Classify and Densify Pavement Distresses

3. 语义分割 (Semantic Segmentation) 🎨

Crack500 / Cracktree200
- 获取链接: https://github.com/fyangneil/pavement-crack-detection.git
- 相关论文: Feature pyramid and hierarchical boosting network for pavement crack detection & CrackTree: Automatic crack detection from pavement images
AsphaltCrack300 / CFD
- 获取链接: https://github.com/guoguolord/CrackDataset
- 相关论文: Automatic road crack detection using random structured forests
GAPs384 / AEL
- 获取链接: https://github.com/fyangneil/pavement-crack-detection.git
- 相关论文: Feature pyramid and hierarchical boosting network… & Automatic Crack Detection on Two-Dimensional Pavement Images…
GAPs 10m
- 获取链接: https://www.tu-ilmenau.de/neurob/data-sets-code/gaps/
- 相关论文: Road Surface Segmentation - Pixel-Perfect Distress and Object Detection for Road Assessment
UDTIRI
- 获取链接: https://www.kaggle.com/datasets/jiahangli617/udtiri
- 相关论文: UDTIRI: An Online Open-Source Intelligent Road Inspection Benchmark Suite

4. 异常与双模态分割

⚠️ 异常分割 - IPAS, AS-Crack500, AS-GAPs
- 获取链接: https://github.com/Yaan-Wang/Pavement-Defect-Datasets.git
- 相关论文: Normal-knowledge-based pavement defect segmentation…
📷 双模态 - Pothole
- 获取链接: https://github.com/ruirangerfan/stereo_pothole_datasets.git
- 相关论文: Pothole detection based on disparity transformation and road surface modeling
📷 双模态 - Pothole-600
- 获取链接: https://sites.google.com/view/pothole-600
- 相关论文: We learn better road pothole detection: from attention aggregation to adversarial domain adaptation
📷 双模态 - FIND
- 获取链接: https://zenodo.org/records/6383044
- 相关论文: Fused image dataset for convolutional neural network-based crack detection