[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-Charmve--Surface-Defect-Detection":3,"tool-Charmve--Surface-Defect-Detection":61},[4,18,26,36,44,53],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":17},4358,"openclaw","openclaw\u002Fopenclaw","OpenClaw 是一款专为个人打造的本地化 AI 助手，旨在让你在自己的设备上拥有完全可控的智能伙伴。它打破了传统 AI 助手局限于特定网页或应用的束缚，能够直接接入你日常使用的各类通讯渠道，包括微信、WhatsApp、Telegram、Discord、iMessage 等数十种平台。无论你在哪个聊天软件中发送消息，OpenClaw 都能即时响应，甚至支持在 macOS、iOS 和 Android 设备上进行语音交互，并提供实时的画布渲染功能供你操控。\n\n这款工具主要解决了用户对数据隐私、响应速度以及“始终在线”体验的需求。通过将 AI 部署在本地，用户无需依赖云端服务即可享受快速、私密的智能辅助，真正实现了“你的数据，你做主”。其独特的技术亮点在于强大的网关架构，将控制平面与核心助手分离，确保跨平台通信的流畅性与扩展性。\n\nOpenClaw 非常适合希望构建个性化工作流的技术爱好者、开发者，以及注重隐私保护且不愿被单一生态绑定的普通用户。只要具备基础的终端操作能力（支持 macOS、Linux 及 Windows WSL2），即可通过简单的命令行引导完成部署。如果你渴望拥有一个懂你",349277,3,"2026-04-06T06:32:30",[13,14,15,16],"Agent","开发框架","图像","数据工具","ready",{"id":19,"name":20,"github_repo":21,"description_zh":22,"stars":23,"difficulty_score":10,"last_commit_at":24,"category_tags":25,"status":17},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,"2026-04-05T11:01:52",[14,15,13],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":32,"last_commit_at":33,"category_tags":34,"status":17},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",142651,2,"2026-04-06T23:34:12",[14,13,35],"语言模型",{"id":37,"name":38,"github_repo":39,"description_zh":40,"stars":41,"difficulty_score":32,"last_commit_at":42,"category_tags":43,"status":17},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107888,"2026-04-06T11:32:50",[14,15,13],{"id":45,"name":46,"github_repo":47,"description_zh":48,"stars":49,"difficulty_score":32,"last_commit_at":50,"category_tags":51,"status":17},4721,"markitdown","microsoft\u002Fmarkitdown","MarkItDown 是一款由微软 AutoGen 团队打造的轻量级 Python 工具，专为将各类文件高效转换为 Markdown 格式而设计。它支持 PDF、Word、Excel、PPT、图片（含 OCR）、音频（含语音转录）、HTML 乃至 YouTube 链接等多种格式的解析，能够精准提取文档中的标题、列表、表格和链接等关键结构信息。\n\n在人工智能应用日益普及的今天，大语言模型（LLM）虽擅长处理文本，却难以直接读取复杂的二进制办公文档。MarkItDown 恰好解决了这一痛点，它将非结构化或半结构化的文件转化为模型“原生理解”且 Token 效率极高的 Markdown 格式，成为连接本地文件与 AI 分析 pipeline 的理想桥梁。此外，它还提供了 MCP（模型上下文协议）服务器，可无缝集成到 Claude Desktop 等 LLM 应用中。\n\n这款工具特别适合开发者、数据科学家及 AI 研究人员使用，尤其是那些需要构建文档检索增强生成（RAG）系统、进行批量文本分析或希望让 AI 助手直接“阅读”本地文件的用户。虽然生成的内容也具备一定可读性，但其核心优势在于为机器",93400,"2026-04-06T19:52:38",[52,14],"插件",{"id":54,"name":55,"github_repo":56,"description_zh":57,"stars":58,"difficulty_score":10,"last_commit_at":59,"category_tags":60,"status":17},4487,"LLMs-from-scratch","rasbt\u002FLLMs-from-scratch","LLMs-from-scratch 是一个基于 PyTorch 的开源教育项目，旨在引导用户从零开始一步步构建一个类似 ChatGPT 的大型语言模型（LLM）。它不仅是同名技术著作的官方代码库，更提供了一套完整的实践方案，涵盖模型开发、预训练及微调的全过程。\n\n该项目主要解决了大模型领域“黑盒化”的学习痛点。许多开发者虽能调用现成模型，却难以深入理解其内部架构与训练机制。通过亲手编写每一行核心代码，用户能够透彻掌握 Transformer 架构、注意力机制等关键原理，从而真正理解大模型是如何“思考”的。此外，项目还包含了加载大型预训练权重进行微调的代码，帮助用户将理论知识延伸至实际应用。\n\nLLMs-from-scratch 特别适合希望深入底层原理的 AI 开发者、研究人员以及计算机专业的学生。对于不满足于仅使用 API，而是渴望探究模型构建细节的技术人员而言，这是极佳的学习资源。其独特的技术亮点在于“循序渐进”的教学设计：将复杂的系统工程拆解为清晰的步骤，配合详细的图表与示例，让构建一个虽小但功能完备的大模型变得触手可及。无论你是想夯实理论基础，还是为未来研发更大规模的模型做准备",90106,"2026-04-06T11:19:32",[35,15,13,14],{"id":62,"github_repo":63,"name":64,"description_en":65,"description_zh":66,"ai_summary_zh":66,"readme_en":67,"readme_zh":68,"quickstart_zh":69,"use_case_zh":70,"hero_image_url":71,"owner_login":72,"owner_name":73,"owner_avatar_url":74,"owner_bio":75,"owner_company":76,"owner_location":77,"owner_email":78,"owner_twitter":79,"owner_website":80,"owner_url":81,"languages":82,"stars":98,"forks":99,"last_commit_at":100,"license":101,"difficulty_score":10,"env_os":102,"env_gpu":103,"env_ram":103,"env_deps":104,"category_tags":107,"github_topics":108,"view_count":32,"oss_zip_url":79,"oss_zip_packed_at":79,"status":17,"created_at":120,"updated_at":121,"faqs":122,"releases":123},4834,"Charmve\u002FSurface-Defect-Detection","Surface-Defect-Detection","📈 目前最大的工业缺陷检测数据库及论文集 Constantly summarizing open source dataset and critical papers in the field of surface defect research which are of great importance.  ","Surface-Defect-Detection 是一个专注于工业表面缺陷检测领域的开源资源库，旨在汇集该方向最重要的公开数据集与核心学术论文。随着机器视觉在 3C、汽车、半导体等制造业中逐步取代人工质检，如何精准识别微小瑕疵成为关键挑战。本项目正是为了解决这一痛点而生，它系统整理了自 2017 年以来的关键研究成果，并针对“小样本训练难”和“实时检测要求高”等行业共性难题提供了丰富的数据支持与理论参考。\n\n资源库不仅涵盖了如 NEU-CLS 钢材表面缺陷等经典数据集，还提供了便捷的下载渠道，帮助使用者快速构建训练环境。对于任务需求，它清晰地将缺陷检测划分为分类（是什么）、定位（在哪里）和分割（有多少）三个层级，便于用户按需探索。无论是从事算法研发的工程师、深耕计算机视觉的研究人员，还是希望了解工业 AI 落地场景的技术决策者，都能从中获得极具价值的素材。通过持续更新的高质量内容，Surface-Defect-Detection 致力于降低行业入门门槛，推动深度学习技术在工业质检场景中的高效应用。","\u003Cdiv align=\"right\">\n  English | \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Fblob\u002Fmaster\u002FReadmeChinese.md\">简体中文\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n# Surface Defect Detection: Dataset & Papers \u003Csup>📌\u003C\u002Fsup>\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGitHub-@Charmve-000000.svg?logo=GitHub\" alt=\"GitHub\" target=\"_blank\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fcharmve.github.io\u002Fcomputer-vision-in-action\u002F\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F计算机视觉实战-简体中文-000000.svg?logo=GitBook\" alt=\"Computer Vision in Action\">\u003C\u002Fa>\n[![License](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002FCharmve\u002FSurface-Defect-Detection)](LICENSE)\n\u003Ca href=\"https:\u002F\u002Fopencollective.com\u002Fsurfacedd\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOpenCollective-Sponsor-000000.svg?logo=OpenCollective&color=purple\" alt=\"Open Collective\">\u003C\u002Fa>\n[![Forks](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fforks\u002FCharmve\u002FSurface-Defect-Detection?style=social)](https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Fedit\u002Fmaster\u002FREADME.md)\n[![Stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FCharmve\u002FSurface-Defect-Detection?style=social)](https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Fedit\u002Fmaster\u002FREADME.md)\n\n\u003Cp>📈 Constantly summarizing open source dataset and critical papers in the field of surface defect research which are of great importance.\nImportant critical papers from year 2017 have been collected and compiled, which can be viewed in the \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ftree\u002Fmaster\u002FPapers\">:open_file_folder: [\u003Cb>\u003Ci>Papers\u003C\u002Fi>\u003C\u002Fb>]\u003C\u002Fa> folder. 🐋 \u003C\u002Fp>\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_7709cc5e712d.png\">\u003C\u002Fdiv>\n\u003Cbr>\n\n\u003Cp align=\"center\">\n  Dataset download: \u003Ccode>\u003Cimg height=\"20\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_62fa12c161e3.png\" alt=\"Google Drive\" title=\"Google Drive\">\u003C\u002Fcode> \u003Ca href=\"https:\u002F\u002Fdrive.google.com\u002Fdrive\u002Ffolders\u002F1q7lirc_yQBXxUSECwX1UvV1TS4eioFm8\">Google Drive\u003C\u002Fa>\n   | \n  \u003Ccode>\u003Cimg height=\"20\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_f7a861308b7e.png\" alt=\"Baidu Cloud\" title=\"Baidu Cloud\">\u003C\u002Fcode> \u003Ca href=\"https:\u002F\u002Fpan.baidu.com\u002Fs\u002F1GWQ_acTF5BnJgpJRSw8BKA\">百度云盘\u003C\u002Fa>  \u003Ccode>o7p5\u003C\u002Fcode>\n\u003C\u002Fp>\n\n## Introduction\n\n\u003Cp>At present, surface defect equipment based on machine vision has widely replaced artificial visual inspection in various industrial fields, including 3C, automobiles, home appliances, machinery manufacturing, semiconductors and electronics, chemical, pharmaceutical, aerospace, light industry and other industries. Traditional surface defect detection methods based on machine vision often use conventional image processing algorithms or artificially designed features plus classifiers. Generally speaking, imaging schemes are usually designed by using the different properties of the inspected surface or defects. A reasonable imaging scheme helps to obtain images with uniform illumination and clearly reflect the surface defects of the object. In recent years, many defect detection methods based on deep learning have also been widely used in various industrial scenarios.\u003C\u002Fp>\n\n\u003Cp>Compared with the clear classification, detection and segmentation tasks in computer vision, the requirements for defect detection are very general. In fact, its requirements can be divided into three different levels: \"what is the defect\" (\u003Cstrong>classification\u003C\u002Fstrong>), \"where is the defect\" (\u003Cstrong>positioning\u003C\u002Fstrong>) and \"How many defects are\" (\u003Cstrong>split\u003C\u002Fstrong>).\u003C\u002Fp>\n\n\u003Cdiv align=\"center\">\n\n*** 本项目会持续更新，右上角收藏防丢失 Star :star: ~ ***\n\n\u003Cb>Star anti-lost\u003C\u002Fb>\n\n\u003Ci>喜欢这个项目吗？请考虑 :heart: 赞助本项目 以帮助长期维护！\u003C\u002Fi>\n\n\u003C\u002Fdiv>\n\n# Table of Contents\n\n- [Introduction](#introduction)\n- [Key Issues](#1-key-issues-in-surface-defect-detection)\n  - [Small Sample Problem](#1small-sample-problem)\n  - [Real-time Problem](#2real-time-problem)\n- [Common Datasets](#2-common-datasets-for-industrial-surface-defect-detection)\n  - [Steel Surface: NEU-CLS](#1steel-surface-neu-cls)\n  - [Kaggle - Severstal: Steel Defect Detection](#kaggle---severstal-steel-defect-detection)\n  - [Solar Panels: elpv-dataset](#2solar-panels-elpv-dataset)\n  - [Metal Surface: KolektorSDD](#3metal-surface-kolektorsdd)\n  - [PCB Inspection: DeepPCB](#4pcb-inspection-deeppcb)\n  - [Fabric Defects Dataset: AITEX](#5fabric-defects-dataset-aitex)\n  - [Fabric Defect Dataset (Tianchi)](#6fabric-defect-dataset-tianchi)\n  - [Aluminium Profile Surface Defect Dataset（Tianchi）](#7aluminium-profile-surface-defect-datasettianchi)\n  - [Weakly Supervised Learning for Industrial Optical Inspection（DAGM 2007）](#8weakly-supervised-learning-for-industrial-optical-inspectiondagm-2007)\n  - [Cracks on the Surface of Construction](#9cracks-on-the-surface-of-the-construction)\n  - [Magnetic Tile Dataset](#10magnetic-tile-dataset)\n  - [RSDDs: Rail Surface Defect Datasets](#11rsdds-rail-surface-defect-datasets)\n  - [Kylberg Texture Dataset v.1.0](#12kylberg-texture-dataset-v10)\n  - [Repeat the Background Texture Dataset: KTH-TIPS](#13KTH-TIPS-database)\n  - [Escalator Step Defect Dataset](#14Escalator-Step-Defect-Dataset) \n  - [Transmission Line Insulator Dataset](#15Transmission-Line-Insulator-Dataset)\n  - [MVTEC ITODD](#16MVTEC-ITODD)\n  - [BSData](#17bsdata---dataset-for-instance-segmentation-and-industrial-wear-forecasting)\n  - [GID: The Gear Inspection Dataset](#18the-gear-inspection-dataset)\n  - [AeBAD aircraft engine blade anomaly detection](#19AeBAD-aircraft-engine-blade-anomaly-detection)\n  - [BeanTech Anomaly Detection(BTAD)](#20BeanTech-Anomaly-Detection(BTAD))\n- [More Inventory](#3-more-inventory-of-the-best-data-set-sources)\n- [Papers](#4-surface-defect-detection-papers)\n- [Acknowledgements](#acknowledgements)\n- [Download](#download)\n- [Notification](#notification)\n- [Community](#-community)\n\n\n## 1. Key Issues in Surface Defect Detection\n\n### 1）Small Sample Problem\n\n\u003Cp>The current deep learning methods are widely used in various computer vision tasks, and surface defect detection is generally regarded as its specific application in the industrial field. In traditional understanding, the reason why deep learning methods cannot be directly applied to surface defect detection is because in a real industrial environment, there are too few industrial defect samples that can be provided.\u003C\u002Fp>\n\n\u003Cp>Compared with the more than 14 million sample data in the ImageNet dataset, the most critical problem faced in surface defect detection is \u003Cb>small sample problem\u003C\u002Fb>. In many real industrial scenarios, there are even only a few or dozens of defective images. In fact, for the small sample problem which is one of the key problems in industrial surface defect detection, there are currently 4 different solutions:\u003C\u002Fp>\n\n\n\u003Cb>- Data Amplification and Generation\u003C\u002Fb>\n\u003Cp> The most commonly used defect image expansion method is to use multiple image processing operations such as mirroring, rotation, translation, distortion, filtering, and contrast adjustment on the original defect samples to obtain more samples. Another more common method is data synthesis, where individual defects are often fused and superimposed on normal (non-defective) samples to form defective samples.\u003C\u002Fp>\n\n\u003Cb>- Network Pre-training and Transfer Learning\u003C\u002Fb>\n\u003Cp>Generally speaking, using small samples to train deep learning networks can easily lead to \u003Cstrong>overfitting\u003C\u002Fstrong>, so methods based on pre-training networks or transfer learning are currently one of the most commonly used methods for samples.\u003C\u002Fp>\n\n\n\u003Cb>- Reasonable Network Structure Design\u003C\u002Fb>\n\u003Cp>The need for samples can also be greatly reduced by designing a reasonable network structure. Based on the compressed sampling theorem to compress and expand small sample data, we use CNN to directly classify the compressed sampling data features. Compared with the original image input, compressing the input can greatly reduce the network's demand for samples. In addition, the surface defect detection method based on the twin network can also be regarded as a special network design, which can greatly reduce the sample requirement.\u003C\u002Fp>\n\n\n\u003Cb>- Unsupervised or Semi-supervised Method\u003C\u002Fb>\n\nIn the unsupervised model, only normal samples are used for training, so there is no need for defective samples. The semi-supervised method can use unlabeled samples to solve the network training problem in the case of small samples.\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 2）Real-time Problem\n\n\u003Cp>The defect detection methods based on deep learning include three main links in industrial applications: \u003Cb>data annotation\u003C\u002Fb>, \u003Cb>model training\u003C\u002Fb>, and \u003Cb>model inference\u003C\u002Fb>. Real-time in actual industrial applications pays more attention to model inference. At present, most defect detection methods are concentrated in the accuracy of classification or recognition, little attention is paid to the efficiency of model inference. There are many methods for accelerating the model, such as model weighting and model pruning. In addition, although the existing deep learning model uses GPU as a general-purpose computing unit(GPGPU), with the development of technology, it is believed that FPGA will become an attractive alternative.\u003C\u002Fp>\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n## 2. Common Datasets for Industrial Surface Defect Detection\n\n### 1）Steel Surface: NEU-CLS\n\nNEU-CLS can be used for classification and positioning tasks.\n\n- :x: Official Link：http:\u002F\u002Ffaculty.neu.edu.cn\u002Fyunhyan\u002FNEU_surface_defect_database.html \n\n\u003Cb> latest access 🔗  - ([#16](https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Fissues\u002F16)) \u003C\u002Fb>\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_046c89cae2a0.png\">\u003C\u002Fdiv>\n\n\u003Cp>The surface defect dataset released by Northeastern University (NEU) collects six typical surface defects of hot-rolled steel strips, namely rolling scale (RS), plaque (Pa), cracking (Cr), pitting surface (PS), inclusions (In) and scratches (Sc). The dataset includes 1,800 grayscale images, six different types of typical surface defects each of which contains 300 samples. For defect detection tasks, the dataset provides annotations that indicate the category and location of the defect in each image. For each defect, the yellow box is the border indicating its location, and the green label is the category score.\u003C\u002Fp>\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_6bc85b694020.png\">\u003C\u002Fdiv>\n\n### Kaggle - Severstal: Steel Defect Detection\n\n\u003Cimg align=\"right\" alt=\"Severstal: Steel Defect Detection\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_6a88fc5444d0.png\" width=\"150\" title=\"Severstal: Steel Defect Detection\">\n\nSeverstal is leading the charge in efficient steel mining and production. They believe the future of metallurgy requires development across the economic, ecological, and social aspects of the industry—and they take corporate responsibility seriously. The company recently created the country’s largest industrial data lake, with petabytes of data that were previously discarded. Severstal is now looking to machine learning to improve automation, increase efficiency, and maintain high quality in their production.\n\nhttps:\u002F\u002Fwww.kaggle.com\u002Fc\u002Fseverstal-steel-defect-detection\n\n\u003Cbr>\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 2）Solar Panels: elpv-dataset\n\n\u003Cp>A dataset of functional and defective solar cells extracted from EL images of solar modules.\u003C\u002Fp>\n\n- 🔗 link：https:\u002F\u002Fgithub.com\u002Fzae-bayern\u002Felpv-dataset\n\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_cea7ade637de.png\">\u003C\u002Fdiv>\n\u003Cbr>\n\nThe dataset contains 2,624 samples of 300x300 pixels 8-bit grayscale images of functional and defective solar cells with varying degree of degradations extracted from 44 different solar modules. The defects in the annotated images are either of intrinsic or extrinsic type and are known to reduce the power efficiency of solar modules.\n\nAll images are normalized with respect to size and perspective. Additionally, any distortion induced by the camera lens used to capture the EL images was eliminated prior to solar cell extraction.\n\n\u003Cbr>\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 3）Metal Surface: KolektorSDD\n\nThe dataset is constructed from images of defected electrical commutators that were provided and annotated by Kolektor Group. Specifically, microscopic fractions or cracks were observed on the surface of the plastic embedding in electrical commutators. The surface area of each commutator was captured in eight non-overlapping images. The images were captured in a controlled environment.\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_7709cc5e712d.png\">\u003C\u002Fdiv>\n\u003Cbr>\n\n- Official Link:https:\u002F\u002Fwww.vicos.si\u002FDownloads\u002FKolektorSDD\n\n- Download Link：https:\u002F\u002Fpan.baidu.com\u002Fshare\u002Finit?surl=HSzHC1ltHvt1hSJh_IY4Jg (password：``1zlb``)\n\n- Implementation： https:\u002F\u002Fgithub.com\u002Fskokec\u002Fsegdec-net-jim2019\n\nThe dataset consists of:\n\n- 50 physical items (defected electrical commutators)\n- 8 surfaces per item\n- Altogether 399 images:\u003Cbr>\n-- 52 images of visible defect\u003Cbr>\n-- 347 images without any defect\n- Original images of sizes:\u003Cbr>\n-- width: 500 px\u003Cbr>\n-- height: from 1240 to 1270 px\n- For training and evaluation images should be resized to 512 x 1408 px\n\nFor each item the defect is only visible in at least one image, while two items have defects on two images, which means there were 52 images where the defects are visible. The remaining 347 images serve as negative examples with non-defective surfaces.\n\n\u003Cbr>\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 4）PCB Inspection: DeepPCB\n\n- 🔗 Download Link：https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ftree\u002Fmaster\u002FDeepPCB\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_27e90b59aed9.jpg\" width=\"375\" style=\"margin:20\">\n     \n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_e14cd5711adf.jpg\" width=\"375\" style=\"margin:20\"> \n \u003C\u002Fdiv>\n\u003Cdiv align=center>\n an example of the tested image \n                                        \n the corresponding template image\n \u003C\u002Fdiv>\n\u003Cp align=center>Figure 1. PCB Inspection Dataset.\u003C\u002Fp>\n\u003Cbr>\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 5）Fabric Defects Dataset: AITEX\n\n- 🔗 Download Link：https:\u002F\u002Fpan.baidu.com\u002Fs\u002F1cfC4Ll5QlnwN5RTuSZ6b7w (password：``b9uy``)\n\n\nThis dataset consists of 245 4096x256 pixel images with seven different fabric structures. There are 140 non-defect images in the dataset, 20 of each type of fabric. In addition, there are 105 images of different types of fabric defects (12 types) common in the textile industry. The image size allows users to use different window sizes, thereby the number of samples can be increased. The online dataset also contains segmentation masks of all defective images, so that white pixels represent defective areas and the remaining pixels are black.\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_e52fc5e40d38.png\">\u003C\u002Fdiv>\n\n\u003Cbr>\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 6）Fabric Defect Dataset (Tianchi)\n\n- 🔗 Download Link：https:\u002F\u002Fpan.baidu.com\u002Fs\u002F1LMbujxvr5iB3SwjFGYHspA (password：``gat2``)\n\n\nIn the actual production process of cloth, due to the influence of various factors, defects such as stains, holes, lint, etc. will occur. In order to ensure the quality of the product, the cloth needs to be inspected for defects. \n\nFabric defect inspection is an important part of the textile industry's production and quality management. At present, manual inspection is susceptible to subjective factors and lacks consistency, and inspection personnel working for a long time under strong light has a great impact on vision. Due to the wide variety of fabric defects, various morphological changes, and the difficulty of observation and recognition, the intelligent detection of fabric defects has been a technical bottleneck that has plagued the industry for many years. \n\nThis dataset covers all kinds of important defects in fabrics in the textile industry, and each picture contains one or more defects. The data includes two types of plain cloth and patterned cloth. Among them, about 8000 pieces of plain cloth data are used for preliminary matches, and about 12,000 pieces of patterned cloth data are used for semi-finals.\n\n\u003Cbr>\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 7）Aluminium Profile Surface Defect Dataset（Tianchi）\n\n- 🔗 Download Link：https:\u002F\u002Ftianchi.aliyun.com\u002Fcompetition\u002Fentrance\u002F231682\u002Finformation\n\nDue to the influence of various factors in the actual production process of aluminum profile, the surface of the aluminum profile will have cracks, peeling, scratches and other defects, which will seriously affect the quality of the aluminum profile. To ensure product quality, manual visual inspection is required. However, the surface of the aluminum profile itself contains textures, which are not highly distinguishable from defects. \n\nTraditional manual visual inspection methods have many shortcomings, which are very laborious, cannot accurately judge surface defects in time, and have difficult to control the efficiency of quality inspection. In recent years, deep learning has made rapid progress in image recognition and other fields. Aluminum profile manufacturers are eager to use the latest AI technology to innovate the existing quality inspection process, automatically complete quality inspection tasks, reduce the incidence of missed inspections, and improve product quality. AI technology, especially deep learning, makes aluminum profile product production managers completely free from the inability to fully grasp the state of product surface quality. \n\nIn the dataset of the competition, there are 10,000 pieces of monitoring image data from aluminum profiles with defects in actual production, and each image contains one or more defects. The sample image for machine learning will clearly identify the type of defect contained in the image.\n\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_32dbe72ffff9.png\">\u003C\u002Fdiv>\n\u003Cbr>\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 8）Weakly Supervised Learning for Industrial Optical Inspection（DAGM 2007） \n\n- 🔗 Download Link：https:\u002F\u002Fhci.iwr.uni-heidelberg.de\u002Fnode\u002F3616\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_23614825725f.png\">\u003C\u002Fdiv>\n\n\u003Cbr>\n\nDataset introduction:\n\n- Mainly aimed at miscellaneous defects on textured backgrounds.\n\n- Training data with weaker supervision.\n\n- Contains ten data sets, the first six are training data sets, and the last four are test data sets.\n\n- Each dataset contains 1000 \"non-defective\" images and 150 \"defective\" images saved in grayscale 8-bit PNG format. Each data set is generated by a different texture model and defect model.\n\n- The background texture of the \"No Defect\" image shows no defect, and the background texture of the \"No Defect\" image has exactly one marked defect.\n\n- All datasets have been randomly divided into training and testing sub-data sets of equal size.\n\n- Weak labels are represented by ellipses, which roughly indicate the defect area.         \n\n\u003Cbr>\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 9）Cracks on the Surface of the Construction\n\nCrackForest Dataset is an annotated road crack image database which can reflect urban road surface condition in general.\n\n- Github Link：https:\u002F\u002Fgithub.com\u002Fcuilimeng\u002FCrackForest-dataset \n\n- Download link：https:\u002F\u002Fpan.baidu.com\u002Fs\u002F1108j5QbDr7T3XQvDxAzVpg (password：``jajn``)\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_f7151187a678.png\">\u003C\u002Fdiv>\n\n\u003Cbr>\n\n\u003Cp align=center>Figure 2. Cracks on the Bridge(left) and Cracks on the Road Surface.\u003C\u002Fp>\n\n- \u003Cb>Bridge cracks\u003C\u002Fb>. There are 2688 images of bridge crack without pixel-level ground truth. From the authors \"Liangfu Li, Weifei Ma, Li Li, Xiaoxiao Gao\". Files can be reached by visiting https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ftree\u002Fmaster\u002FBridge_Crack_Image.\n\n- \u003Cb>Crack on road surface\u003C\u002Fb>. From Shi Yong, and Cui Limeng and Qi Zhiquan and Meng Fan and Chen Zhensong. Original dataset can be reached at https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ftree\u002Fmaster\u002FCrackForest. We extract the image files of the pixel level ground truth.\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 10）Magnetic Tile Dataset\n\nMagnetic tile dataset by githuber: abin24, which can be downloaded from [https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ftree\u002Fmaster\u002FMagnetic-Tile-Defect](https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ftree\u002Fmaster\u002FMagnetic-Tile-Defect), which was used in their paper \"Surface defect saliency of magnetic tile\", the paper can be reach by [here](https:\u002F\u002Flink.springer.com\u002Farticle\u002F10.1007\u002Fs00371-018-1588-5) or [here](https:\u002F\u002Fieeexplore.ieee.org\u002Fdocument\u002F8560423)\n\n![dataset](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_543b47e40714.jpg) \n\n\u003Cp align=center>Figure 3. An overview of our dataset.\u003C\u002Fp>\n\nThis is also the datasets of the paper \"Saliency of magnetic tile surface defects\". \nThe images of 6 common magnetic tile defects were collected, and their pixel level ground-truth were labeled.\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 11）RSDDs: Rail Surface Defect Datasets\n\nThe RSDDs dataset contains two types of datasets: the first is a type I RSDDs dataset captured from the fast lane, which contains 67 challenging images. The second is a Type II RSDDs dataset captured from a normal\u002Fheavy transportation track, which contains 128 challenging images.\n\nEach image of the two data sets contains at least one defect, and the background is complex and noisy.\n\nThese defects in the RSDDs dataset have been marked by professional human observers in the field of track surface inspection.\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_f3b15d469e88.jpg\">\u003C\u002Fdiv>\n\u003Cbr>\n\n- Official Link：http:\u002F\u002Ficn.bjtu.edu.cn\u002FVisint\u002Fresources\u002FRSDDs.aspx\n\n- Download Link：https:\u002F\u002Fpan.baidu.com\u002Fshare\u002Finit?surl=svsnqL0r1kasVDNjppkEwg (password：``nanr``)\n\n\u003Cbr>\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 12）Kylberg Texture Dataset v.1.0 \n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_d947bf4fd5d0.png\">\u003C\u002Fdiv>\n\u003Cp align=center>Figure 4. Example patches from each one of the 28 texture classes.\u003C\u002Fp>\n\nShort description\n- 28 texture classes, see Figure 4.\n- 160 unique texture patches per class. (Alternative dataset with 12 rotations per each original patch, 160*12=1920 texture patches per class).\n- Texture patch size: 576x576 pixels.\n- File format: Lossless compressed 8 bit PNG.\n- All patches are normalized with a mean value of 127 and a standard deviation of 40.\n- One directory per texture class.\n- Files are named as follows: ``blanket1-d-p011-r180.png``, where ``blanket1`` is the class name, ``d`` original image sample number (possible values are a, b, c, or d), ``p011`` is patch number 11, ``r180`` patch rotated 180 degrees.\n\n🔗 Offical Link: http:\u002F\u002Fwww.cb.uu.se\u002F~gustaf\u002Ftexture\u002F\n\u003Cbr>\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 13）KTH-TIPS database\n\nRepeat the background texture data set, the sample picture is as follows\n\n\u003Cdiv align=center>\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_2483718ef397.png\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_d5f956cace04.png\">\n\u003C\u002Fdiv>\n\n- Offical Link:https:\u002F\u002Fwww.nada.kth.se\u002Fcvap\u002Fdatabases\u002Fkth-tips\u002Fdownload.html\n\n- Download Link：\n\n  - dataset1：https:\u002F\u002Fpan.baidu.com\u002Fs\u002F173h8V66yRmtVo5rc2P7J4A\n\n  - dataset2：https:\u002F\u002Fpan.baidu.com\u002Fs\u002F1dXFKn6v2PV5QS9m8gWlifA\n\n\u003Cbr>\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n\n### 14）Escalator Step Defect Dataset \n\n🔗 Offical Link：https:\u002F\u002Faistudio.baidu.com\u002Faistudio\u002Fdatasetdetail\u002F44820\n\n\u003Cbr>\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 15）Transmission Line Insulator Dataset\n\nIn the data set, ``Normal_Insulators`` contains 600 insulator images captured by drones. ``Defective_Insulators`` contains defective insulators, and the number of defective images of insulators is 248. The data set includes data sets and labels.\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_a13a48c416ed.png\">\u003C\u002Fdiv>\n\n🔗 Offical Link：https:\u002F\u002Fgithub.com\u002FInsulatorData\u002FInsulatorDataSet\n\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 16）MVTEC ITODD\n\nThe **MVTec Industrial 3D Object Detection Dataset (MVTec ITODD)** is a public dataset for 3D object detection and pose estimation with a strong focus on industrial settings and applications.\n\nThe dataset consists of\n\n- 28 objects and 3500 labeled scenes containing instances of these objects\n- Five sensors (two 3D sensors and three grayscale cameras) observing each scene\n\nMore information can be found in [this PDF file](https:\u002F\u002Fwww.mvtec.com\u002Ffileadmin\u002FRedaktion\u002Fmvtec.com\u002Fcompany\u002Fresearch\u002Fdatasets\u002Fmvtec_itodd.pdf) 🔍.\n\n\u003Cdiv align=center>\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_e2380bbc79f1.png\">\n\u003C\u002Fdiv>\n\n🔗 Download link https:\u002F\u002Fwww.mvtec.com\u002Fcompany\u002Fresearch\u002Fdatasets\u002Fmvtec-itodd\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 17）BSData - dataset for Instance Segmentation and industrial Wear Forecasting\n\nThe dataset contains 1104 channel 3 images with 394 image-annotations for the surface damage \ntype “pitting”. The annotations made with the annotation tool [labelme](https:\u002F\u002Fgithub.com\u002Fwkentaro\u002Flabelme), \nare available in ``JSON`` format and hence convertible to VOC and COCO format. All images come from two BSD types. \n\nThe other BSD type is shown on 325 images with two image-sizes. Since all images of this type have been taken with continuous \ntime the degree of soiling is evolving.\n\nAlso, the dataset contains as above mentioned 27 pitting development sequences with every 69 images.\n\n\u003Cp align=center>\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_31159abccd9f.png\">\n  Figure 5. On the left image-examples, on the right associated PNG-Annotations.\n\u003C\u002Fp>\n\n🔗 Offical link https:\u002F\u002Fgithub.com\u002F2Obe\u002FBSData\n\n> Sincerely, thank @Beñat Gartzia for his recommendation and all your attention!\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 18）The Gear Inspection Dataset\n\nThe Gear Inspection Dataset (GID) is a dataset for a competition held by Baidu (China) Co., called the \"National Artificial Intelligence Innovation Application Competition.\" It has two thousand grayscale images with 28575 annotations for three types of defects from a real-world source. Each picture includes defects described in a separate JSON file with the image name, label categories, bounding boxes, and polygons for segmentation. Nevertheless, the tags for labeling categories do not include specific information about their type but only numbers, so spotting their similarities with other related datasets is challenging.\n\n\u003Cp align=center>\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_b9bfa9450a88.jpg\">\n    Figure 6. Examples of validation test images and their labels.\n\u003C\u002Fp>\n\n🔗 Offical link http:\u002F\u002Fwww.aiinnovation.com.cn\u002F#\u002FdataDetail?id=34\n\n- Download Link：\n  - Gear Detection Training Dataset: https:\u002F\u002Fpan.baidu.com\u002Fs\u002F17HoFfBUQGeX7G0ibkPExrw (passwprd: hm7k) \n  - Gear detection A list evaluation dataset: https:\u002F\u002Fpan.baidu.com\u002Fs\u002F157Zf7hcTM78GhXtXI5ySFQ (pass: 2R6K)\n  - Gear detection B list evaluation dataset: https:\u002F\u002Fpan.baidu.com\u002Fs\u002F1OjOZotqlRSvsYLA_qH2nXA (pass: hypd)\n  \n- Mirrors:\n  - Gear Detection Training Dataset: https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1CZo-Ab5BXkTjV-b1-NIFzYMjfJQMl4nG\u002Fview?usp=share_link\n  - Gear detection A list evaluation dataset: https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1-0sSrmhElBseeZWICu77lzTxoOiRD8yG\u002Fview?usp=share_link\n  - Gear detection B list evaluation dataset: N\u002FA.\n  \n Note: The contest dataset is not for commercial use.\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n\n### 19）AeBAD Aircraft Engine Blade Anomaly Detection\n\nDownload link: http:\u002F\u002Fsuo.nz\u002F2IU48P\n\nThe real-world aero-engine blade anomaly detection (AeBAD) data set consists of two sub-data sets: the single blade data set (AeBAD-S) and the blade video anomaly detection data set (AeBAD-V). Compared with existing datasets, AeBAD has the following two characteristics: 1.) The target samples are not aligned and at different scales. 2.) There is a domain shift in the distribution of normal samples in the test set and training set, where the domain shift is mainly caused by changes in illumination and view.\n\n![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_ef046a4da7cf.png)\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n### 20）BeanTech Anomaly Detection(BTAD)\n\nDownload Link：http:\u002F\u002Fsuo.nz\u002F2JEGEi\n\nThe BTAD (BeanTech Anomaly Detection) dataset is a real-world industrial anomaly dataset. This dataset contains a total of 2830 real-world images of 3 industrial products.\n\n![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_151bd15e2354.png)\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n\u003Cbr>\n\n## 3. More Inventory of the Best Data Set Sources\n\nI have been collecting surface defect detection data sets, but there are still many data sets that have not been collected. For the data sets not collected in this repo, you can go to the following sites to view. \u003Cb>At the same time, everyone is very welcome to share the new data set and become the contributor of this repo.\u003C\u002Fb>\n\n![Contributions welcome](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fcontributing-👐%20Welcome-orange.svg)\n\n|source|url|Recommended|\n|--|--|--|\n|Kaggle|https:\u002F\u002Fwww.kaggle.com\u002Fdatasets|⭐⭐⭐⭐⭐|\n|Paper With Code |https:\u002F\u002Fpaperwithcode.com\u002Fsota|⭐⭐⭐⭐⭐|\n|Registry of Open Data on AWS|https:\u002F\u002Fregistry.opendata.aws|⭐⭐⭐|\n|Microsoft Research Open Data|https:\u002F\u002Fmsropendata.com |⭐⭐⭐|\n|Awesome-public-datasets|https:\u002F\u002Fgithub.com\u002Fawesomedata\u002Fawesome-public-datasets |⭐⭐|\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n\u003Cbr>\n\n## 4. Surface Defect Detection Papers\n\nI have collected some articles on surface defect detection. The main objects to be tested are: defects or abnormal objects such as metal surfaces, LCD screens, buildings, and power lines. The methods are mainly classified method, detection method, reconstruction method and generation method. The electronic version (PDF) of the paper is placed under the file named corresponding to the date in the 'Paper' folder.\n\nGo to \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ftree\u002Fmaster\u002FPapers\">:open_file_folder: [\u003Cb>\u003Ci>Papers\u003C\u002Fi>\u003C\u002Fb>]\u003C\u002Fa>.\n\n\u003Cbr>\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n## Acknowledgements\n\n\u003Cp>You can see this repo now, we should be grateful to the people who originally open sourced the above data set. They have brought great help to our study and research work. The idea of collecting this data set originally came from reading an article on surface defect detection by SFXiang of \"AI算法修炼营(AI_SuanFa)\", which prompted me to organize a more comprehensive data set. The collection of papers comes from a CSDNer named \"庆志的小徒弟\". These papers are only until November 19, and I will continue to be improved after that. \u003Cstrong>Hopefully, feel free to \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Fblob\u002Fmaster\u002FCONTRIBUTING.md\" target=\"_blank\">CONTRIBUTE\u003C\u002Fa>.\u003C\u002Fstrong>\u003C\u002Fp>\n\n\u003Cp>Finally, I want to thank the open source contributors of the above data set again.\u003C\u002Fp>\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n## Download\n- Download ZIP, click [here](https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Farchive\u002Fmaster.zip)\n  \u003Cbr>or run ```git clone https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection.git``` in the terminal\u003Cbr>\n- Chinese Mainland - 百度网盘下载链接 https:\u002F\u002Fpan.baidu.com\u002Fs\u002F122WY8F5VKqm3qMirqebRQw ``提取码:i20n``\n\n👆 [\u003Cb>BACK to Table of Contents\u003C\u002Fb> -->](#table-of-contents)\n\n## Notification\n\nThis work is originally contributed by lots of great man for their paper work or industry application. \u003Cstrong>You can only use this dataset for research purpose.\u003C\u002Fstrong>\n\nIf you have any questions or idea, please let me know :email: yidazhang1@gmail.com\n\n## 🍮 Community\n- Github \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Fdiscussions\" target=\"_blank\">discussions 💬\u003C\u002Fa> or \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Fissues\" target=\"_blank\">issues 💭\u003C\u002Fa>\n\n- QQ Group: 734758251 (password：哈哈哈)\n- WeChat Group ID: Yida_Zhang2\n- Email: yidazhang1#gmail.com \n\n\u003Cbr>\n\n\u003Ca href=\"https:\u002F\u002Fcharmve.github.io\u002Fsponsor.html\">\u003Cimg align=\"left\" alt=\"Go for it!\" src=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002Fcomputer-vision-in-action\u002Fblob\u002Fmain\u002Fres\u002Fui\u002Ffrontpage\u002F2020-sponsors.svg\" height=\"120\" title=\"Do what you like, and do it best!\"\u002F>\u003C\u002Fa>\n\n##    Supporting\n\n      Support this project by becoming a sponsor. Your name and\u002For logo will show up \u003Cb>our homepage\u003C\u002Fb> with a link to your website. 🙏\n\n\u003Cp align=\"center\">\n    \u003Ca href=\"https:\u002F\u002Fcharmve.github.io\u002Fsponsor.html\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fcdn.buymeacoffee.com\u002Fbuttons\u002Fv2\u002Fdefault-red.png\" alt=\"Sponse this project\" width=\"280\" >\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003C!-- Backers\n\n\u003Cdiv align=\"left\">\n    \u003Ca href=\"https:\u002F\u002Fopencollective.com\u002Fsurfacedd#sponsors\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fopencollective.com\u002Fl0cv\u002Fsponsors.svg?width=200\">\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n\n\u003Cbr>\n\u003Cdiv class=\"sponsor-level\">\n    \u003Ch2 align=\"center\">Our Backers\u003C\u002Fh2>\n    \u003Ch4 align=\"center\">\u003Cem>Waiting for you, code with ❤️ ！\u003C\u002Fem>\u003C\u002Fh4>\n    \u003C!-- \u003Ch3>\u003Cem>Waiting for you, Code with ❤️ ！\u003C\u002Fem>\u003C\u002Fh3> --\n    \u003Ca href=\"https:\u002F\u002Fopencollective.com\u002Fsurfacedd#backers\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fopencollective.com\u002Fsurfacedd\u002Fbackers.svg?width=200\">\u003C\u002Fa>\n    \u003Ctable align=\"center\">\n      \u003Ctbody>\n\t \u003Cth>\n\t   \u003Ctd>\n\t\t   date 📅 \n\t   \u003C\u002Ftd>\n\t   \u003Ctd>\n\t\t   boss 👵 \n\t   \u003C\u002Ftd>\n\t   \u003Ctd>\n\t\t   amount 💰\n\t   \u003C\u002Ftd>\n\t   \u003Ctd>\n\t\t   way ♥️\n\t   \u003C\u002Ftd>\n\t   \u003Ctd>\n\t\t   message :speech_balloon:\n\t   \u003C\u002Ftd>\n\t \u003C\u002Fth>\n\t \u003Ctr>\n           \u003Ctd>\n\t\t   01\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   2022.02.06\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   @大巧不工\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   ￥6.66 \n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   微信红包\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t   \u003C\u002Ftd>\n         \u003C\u002Ftr>\n\t \u003Ctr>\n           \u003Ctd>\n\t\t   02\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   2022.02.08\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   @大巧不工\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   ￥6\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   微信红包\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t   \u003C\u002Ftd>\n         \u003C\u002Ftr>\n      \u003C\u002Ftbody>\n    \u003C\u002Ftable>\n    \u003Cbr>\n    \u003Csup>* \u003Cb>Note\u003C\u002Fb>: Because WeChat and Alipay do not support personal interface calls, the information on the sponsorship list is manually added by me, and there may be a delay in updating. Just in case, you can also initiate an Issue stating your corresponding sponsorship and the help you want to get. I will close the corresponding Issue after adding\u002Fsolving.\u003C\u002Fsup>\n\u003C\u002Fdiv>\n\n--->\n\n## Citation\nUse this bibtex to cite this repository:\n```\n@misc{Surface Defect Detection,\n  title={Surface Defect Detection: Dataset and Papers},\n  author={Charmve},\n  year={2020.09},\n  publisher={Github},\n  journal={GitHub repository},\n  howpublished={\\url{https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection}},\n}\n```\n\n## Stargazers over time\n\n[![Stargazers over time](https:\u002F\u002Fstarchart.cc\u002FCharmve\u002FSurface-Defect-Detection.svg)](https:\u002F\u002Fstarchart.cc\u002FCharmve\u002FSurface-Defect-Detection)\n\n\u003Cbr>\n\u003Cp align=\"center\">Feel free to ask any questions, open a PR if you feel something can be done differently!\u003C\u002Fp>\n\u003Ch2 align=\"center\">🌟Star this repository🌟\u003C\u002Fh2>\n\u003Cp align=\"center\">Created by \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\">Charmve\u003C\u002Fa> & \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FMaiweiAI\">maiwei.ai\u003C\u002Fa> Community | Deployed on \u003Ca href=\"https:\u002F\u002Fwww.kaggle.com\u002Fyidazhang07\u002Fbridge-cracks-image\">Kaggle\u003C\u002Fa>\u003C\u002Fp>\n\n\u003Cbr>\n* \u003Ci>Update on Sep 17, 2021 @\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\" target=\"_blank\">Charmve\u003C\u002Fa>, \n    \u003Ca class=\"github-button\"\n        href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\"\n        data-icon=\"octicon-star\" data-show-count=\"true\"\n        aria-label=\"Star Charmve\u002FSurface-Defect-Detection on GitHub\">Star\u003C\u002Fa> \n    and \n    \u003Ca class=\"github-button\"\n        href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ffork\"\n        data-icon=\"octicon-repo-forked\" data-show-count=\"true\"\n        aria-label=\"Fork Charmve\u002FSurface-Defect-Detection on GitHub\">Fork\u003C\u002Fa>\n\u003C\u002Fi>\n","\u003Cdiv align=\"right\">\n  英文 | \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Fblob\u002Fmaster\u002FReadmeChinese.md\">简体中文\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n# 表面缺陷检测：数据集与论文 \u003Csup>📌\u003C\u002Fsup>\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGitHub-@Charmve-000000.svg?logo=GitHub\" alt=\"GitHub\" target=\"_blank\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fcharmve.github.io\u002Fcomputer-vision-in-action\u002F\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F计算机视觉实战-简体中文-000000.svg?logo=GitBook\" alt=\"Computer Vision in Action\">\u003C\u002Fa>\n[![License](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002FCharmve\u002FSurface-Defect-Detection)](LICENSE)\n\u003Ca href=\"https:\u002F\u002Fopencollective.com\u002Fsurfacedd\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOpenCollective-Sponsor-000000.svg?logo=OpenCollective&color=purple\" alt=\"Open Collective\">\u003C\u002Fa>\n[![Forks](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fforks\u002FCharmve\u002FSurface-Defect-Detection?style=social)](https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Fedit\u002Fmaster\u002FREADME.md)\n[![Stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FCharmve\u002FSurface-Defect-Detection?style=social)](https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Fedit\u002Fmaster\u002FREADME.md)\n\n\u003Cp>📈 不断总结表面缺陷研究领域中具有重要意义的开源数据集和关键论文。\n自2017年以来的重要关键论文已被收集并整理，可在 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ftree\u002Fmaster\u002FPapers\">:open_file_folder: [\u003Cb>\u003Ci>Papers\u003C\u002Fi>\u003C\u002Fb>]\u003C\u002Fa> 文件夹中查看。 🐋 \u003C\u002Fp>\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_7709cc5e712d.png\">\u003C\u002Fdiv>\n\u003Cbr>\n\n\u003Cp align=\"center\">\n  数据集下载： \u003Ccode>\u003Cimg height=\"20\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_62fa12c161e3.png\" alt=\"Google Drive\" title=\"Google Drive\">\u003C\u002Fcode> \u003Ca href=\"https:\u002F\u002Fdrive.google.com\u002Fdrive\u002Ffolders\u002F1q7lirc_yQBXxUSECwX1UvV1TS4eioFm8\">Google Drive\u003C\u002Fa>\n   | \n  \u003Ccode>\u003Cimg height=\"20\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_f7a861308b7e.png\" alt=\"Baidu Cloud\" title=\"Baidu Cloud\">\u003C\u002Fcode> \u003Ca href=\"https:\u002F\u002Fpan.baidu.com\u002Fs\u002F1GWQ_acTF5BnJgpJRSw8BKA\">百度云盘\u003C\u002Fa>  \u003Ccode>o7p5\u003C\u002Fcode>\n\u003C\u002Fp>\n\n## 引言\n\n\u003Cp>目前，基于机器视觉的表面缺陷检测设备已广泛取代人工目视检查，应用于多个工业领域，包括3C、汽车、家电、机械制造、半导体与电子、化工、制药、航空航天、轻工等行业。传统的基于机器视觉的表面缺陷检测方法通常采用常规图像处理算法或人工设计特征结合分类器。一般来说，成像方案会根据被检表面或缺陷的不同特性来设计，合理的成像方案有助于获得光照均匀且能清晰反映物体表面缺陷的图像。近年来，许多基于深度学习的缺陷检测方法也在各类工业场景中得到广泛应用。\u003C\u002Fp>\n\n\u003Cp>相较于计算机视觉中的清晰分类、检测和分割任务，表面缺陷检测的需求则更为通用。实际上，其需求可分为三个不同层次：“是什么缺陷”（\u003Cstrong>分类\u003C\u002Fstrong>）、“在哪里缺陷”（\u003Cstrong>定位\u003C\u002Fstrong>）以及“有多少缺陷”（\u003Cstrong>计数\u003C\u002Fstrong>）。\u003C\u002Fp>\n\n\u003Cdiv align=\"center\">\n\n*** 本项目会持续更新，右上角收藏防丢失 Star :star: ~ ***\n\n\u003Cb>Star 防丢\u003C\u002Fb>\n\n\u003Ci>喜欢这个项目吗？请考虑 :heart: 赞助本项目 以帮助长期维护！\u003C\u002Fi>\n\n\u003C\u002Fdiv>\n\n# 目录\n\n- [引言](#introduction)\n- [关键问题](#1-key-issues-in-surface-defect-detection)\n  - [小样本问题](#1small-sample-problem)\n  - [实时性问题](#2real-time-problem)\n- [常用数据集](#2-common-datasets-for-industrial-surface-defect-detection)\n  - [钢铁表面：NEU-CLS](#1steel-surface-neu-cls)\n  - [Kaggle - Severstal：钢铁缺陷检测](#kaggle---severstal-steel-defect-detection)\n  - [太阳能电池板：elpv-dataset](#2solar-panels-elpv-dataset)\n  - [金属表面：KolektorSDD](#3metal-surface-kolektorsdd)\n  - [PCB 检查：DeepPCB](#4pcb-inspection-deeppcb)\n  - [织物缺陷数据集：AITEX](#5fabric-defects-dataset-aitex)\n  - [织物缺陷数据集（天池）](#6fabric-defect-dataset-tianchi)\n  - [铝型材表面缺陷数据集（天池）](#7aluminium-profile-surface-defect-datasettianchi)\n  - [面向工业光学检测的弱监督学习（DAGM 2007）](#8weakly-supervised-learning-for-industrial-optical-inspectiondagm-2007)\n  - [建筑表面裂缝](#9cracks-on-the-surface-of-the-construction)\n  - [磁性瓷砖数据集](#10magnetic-tile-dataset)\n  - [RSDDs：铁路表面缺陷数据集](#11rsdds-rail-surface-defect-datasets)\n  - [Kylberg 纹理数据集 v.1.0](#12kylberg-texture-dataset-v10)\n  - [重复背景纹理数据集：KTH-TIPS](#13KTH-TIPS-database)\n  - [自动扶梯踏板缺陷数据集](#14Escalator-Step-Defect-Dataset) \n  - [输电线路绝缘子数据集](#15Transmission-Line-Insulator-Dataset)\n  - [MVTEC ITODD](#16MVTEC-ITODD)\n  - [BSData](#17bsdata---dataset-for-instance-segmentation-and-industrial-wear-forecasting)\n  - [GID：齿轮检测数据集](#18the-gear-inspection-dataset)\n  - [AeBAD 飞机发动机叶片异常检测](#19AeBAD-aircraft-engine-blade-anomaly-detection)\n  - [BeanTech 异常检测（BTAD）](#20BeanTech-Anomaly-Detection(BTAD))\n- [更多资源](#3-more-inventory-of-the-best-data-set-sources)\n- [论文](#4-surface-defect-detection-papers)\n- [致谢](#acknowledgements)\n- [下载](#download)\n- [通知](#notification)\n- [社区](#-community)\n\n\n## 1. 表面缺陷检测中的关键问题\n\n### 1）小样本问题\n\n\u003Cp>当前深度学习方法已被广泛应用于各类计算机视觉任务中，而表面缺陷检测通常被视为其在工业领域的具体应用之一。按照传统认知，深度学习方法难以直接应用于表面缺陷检测的原因在于，在真实的工业环境中，可用于训练的缺陷样本数量极其有限。\u003C\u002Fp>\n\n\u003Cp>与ImageNet数据集中超过1400万张样本数据相比，表面缺陷检测所面临的核心问题正是\u003Cb>小样本问题\u003C\u002Fb>。在许多实际工业场景中，甚至只有寥寥几张或几十张缺陷图像。事实上，针对这一工业表面缺陷检测中的关键问题，目前主要有四种解决方案：\u003C\u002Fp>\n\n\n\u003Cb>- 数据增强与生成\u003C\u002Fb>\n\u003Cp>最常用的缺陷图像扩充方法是在原始缺陷样本上应用多种图像处理操作，如镜像、旋转、平移、扭曲、滤波和对比度调整等，从而生成更多样本。另一种更为常见的方法是数据合成，即将单个缺陷特征融合并叠加到正常（无缺陷）样本上，以形成带有缺陷的样本。\u003C\u002Fp>\n\n\u003Cb>- 网络预训练与迁移学习\u003C\u002Fb>\n\u003Cp>一般来说，使用少量样本训练深度学习网络容易导致\u003Cstrong>过拟合\u003C\u002Fstrong>,因此基于网络预训练或迁移学习的方法是目前处理小样本问题时最常用的方式之一。\u003C\u002Fp>\n\n\n\u003Cb>- 合理的网络结构设计\u003C\u002Fb>\n\u003Cp>通过设计合理的网络结构，也可以大幅降低对样本的需求。例如，基于压缩采样定理对小样本数据进行压缩与扩展，并利用卷积神经网络直接对压缩后的采样特征进行分类。相较于原始图像输入，压缩输入能够显著减少网络对样本的需求。此外，基于孪生网络的表面缺陷检测方法也可视为一种特殊的网络设计，同样能大幅降低对样本的需求。\u003C\u002Fp>\n\n\n\u003Cb>- 无监督或半监督方法\u003C\u002Fb>\n\n在无监督模型中，仅使用正常样本进行训练，因此无需缺陷样本。而半监督方法则可以在仅有少量标注样本的情况下，利用未标注样本解决网络训练问题。\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 2）实时性问题\n\n\u003Cp>基于深度学习的缺陷检测方法在工业应用中主要包括三个环节：\u003Cb>数据标注\u003C\u002Fb>、\u003Cb>模型训练\u003C\u002Fb>和\u003Cb>模型推理\u003C\u002Fb>。而在实际工业应用中，实时性主要关注的是模型推理阶段。目前，大多数缺陷检测方法更注重分类或识别的准确性，而对模型推理效率的关注相对较少。加速模型运行的方法有很多，例如模型量化和剪枝等。此外，尽管现有的深度学习模型普遍采用GPU作为通用计算单元（GPGPU），但随着技术的发展，FPGA被认为将成为一个极具吸引力的替代方案。\u003C\u002Fp>\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n## 2. 工业表面缺陷检测常用数据集\n\n### 1）钢铁表面：NEU-CLS\n\nNEU-CLS可用于分类和定位任务。\n\n- :x: 官方链接：http:\u002F\u002Ffaculty.neu.edu.cn\u002Fyunhyan\u002FNEU_surface_defect_database.html \n\n\u003Cb> 最新访问 🔗  - ([#16](https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Fissues\u002F16)) \u003C\u002Fb>\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_046c89cae2a0.png\">\u003C\u002Fdiv>\n\n\u003Cp>由东北大学（NEU）发布的表面缺陷数据集收录了热轧钢带的六种典型表面缺陷，分别为氧化铁皮（RS）、斑块（Pa）、裂纹（Cr）、麻点表面（PS）、夹杂物（In）和划痕（Sc）。该数据集包含1,800张灰度图像，每种典型的表面缺陷各有300个样本。对于缺陷检测任务，数据集提供了标注信息，标明了每张图像中缺陷的类别和位置。对于每种缺陷，黄色框表示其边界位置，绿色标签则为类别得分。\u003C\u002Fp>\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_6bc85b694020.png\">\u003C\u002Fdiv>\n\n### Kaggle - Severstal：钢铁缺陷检测\n\n\u003Cimg align=\"right\" alt=\"Severstal：钢铁缺陷检测\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_6a88fc5444d0.png\" width=\"150\" title=\"Severstal：钢铁缺陷检测\">\n\nSeverstal公司在高效钢铁开采和生产方面处于领先地位。他们认为，冶金行业的未来需要在经济、生态和社会等多个层面同步发展，并且高度重视企业社会责任。该公司最近建立了国内最大的工业数据湖，存储了此前被废弃的数PB级数据。如今，Severstal正借助机器学习技术来提升自动化水平、提高生产效率并确保产品质量。\n\nhttps:\u002F\u002Fwww.kaggle.com\u002Fc\u002Fseverstal-steel-defect-detection\n\n\u003Cbr>\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 2）太阳能电池板：elpv-dataset\n\n\u003Cp>从太阳能组件的EL图像中提取的功能性和缺陷性太阳能电池数据集。\u003C\u002Fp>\n\n- 🔗 链接：https:\u002F\u002Fgithub.com\u002Fzae-bayern\u002Felpv-dataset\n\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_cea7ade637de.png\">\u003C\u002Fdiv>\n\u003Cbr>\n\n该数据集包含2,624个样本，均为300×300像素、8位灰度图像，涵盖了来自44个不同太阳能组件的功能性和不同程度退化的缺陷性太阳能电池。标注中的缺陷分为内在型和外在型两类，已知会降低太阳能组件的发电效率。\n\n所有图像均经过尺寸和视角的归一化处理。此外，在提取太阳能电池之前，还去除了拍摄EL图像时相机镜头造成的任何畸变。\n\n\u003Cbr>\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 3）金属表面：KolektorSDD\n\n该数据集由Kolektor集团提供并标注的有缺陷电枢换向器图像构成。具体而言，在电枢换向器的塑料封装表面观察到了微观裂纹或裂缝。每个换向器的表面被分割成八张不重叠的图像进行拍摄，所有图像均在受控环境中采集。\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_7709cc5e712d.png\">\u003C\u002Fdiv>\n\u003Cbr>\n\n- 官方链接：https:\u002F\u002Fwww.vicos.si\u002FDownloads\u002FKolektorSDD\n\n- 下载链接：https:\u002F\u002Fpan.baidu.com\u002Fshare\u002Finit?surl=HSzHC1ltHvt1hSJh_IY4Jg（提取码：``1zlb``）\n\n- 实现代码：https:\u002F\u002Fgithub.com\u002Fskokec\u002Fsegdec-net-jim2019\n\n数据集包含：\n\n- 50个实物样本（有缺陷的电枢换向器）\n- 每个样本8个表面\n- 共计399张图像：\u003Cbr>\n-- 52张可见缺陷的图像\u003Cbr>\n-- 347张无缺陷的图像\n- 原始图像尺寸：\u003Cbr>\n-- 宽度：500像素\u003Cbr>\n-- 高度：1240至1270像素\n- 用于训练和评估的图像应调整为512×1408像素。\n\n对于每个样本，缺陷仅在至少一张图像中可见；其中两个样本在两张图像上有缺陷，因此共有52张图像显示了缺陷。其余347张图像则作为无缺陷的负例。\n\n\u003Cbr>\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 4）PCB检测：DeepPCB\n\n- 🔗 下载链接：https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ftree\u002Fmaster\u002FDeepPCB\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_27e90b59aed9.jpg\" width=\"375\" style=\"margin:20\">\n     \n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_e14cd5711adf.jpg\" width=\"375\" style=\"margin:20\"> \n \u003C\u002Fdiv>\n\u003Cdiv align=center>\n 测试图像示例 \n                                        \n 对应的模板图像\n \u003C\u002Fdiv>\n\u003Cp align=center>图1. PCB检测数据集。\u003C\u002Fp>\n\u003Cbr>\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 5）织物缺陷数据集：AITEX\n\n- 🔗 下载链接：https:\u002F\u002Fpan.baidu.com\u002Fs\u002F1cfC4Ll5QlnwN5RTuSZ6b7w（提取码：``b9uy``）\n\n\n该数据集包含245张4096×256像素的图像，涵盖了七种不同的织物结构。数据集中有140张无缺陷图像，每种织物各20张。此外，还有105张不同类型的织物缺陷图像（共12种），这些缺陷是纺织行业中常见的类型。图像的高分辨率允许用户使用不同的窗口大小，从而增加样本数量。在线数据集还包含了所有缺陷图像的分割掩膜，其中白色像素表示缺陷区域，其余部分为黑色。\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_e52fc5e40d38.png\">\u003C\u002Fdiv>\n\n\u003Cbr>\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 6）织物缺陷数据集（天池）\n\n- 🔗 下载链接：https:\u002F\u002Fpan.baidu.com\u002Fs\u002F1LMbujxvr5iB3SwjFGYHspA（提取码：``gat2``）\n\n\n在实际的布料生产过程中，由于各种因素的影响，会出现污渍、破洞、毛絮等缺陷。为了保证产品质量，必须对布料进行缺陷检测。\n\n织物缺陷检测是纺织工业生产和质量管理中的重要环节。目前，人工检测容易受到主观因素影响且一致性较差，长时间在强光下工作的检测人员视力也会受到很大影响。由于织物缺陷种类繁多、形态变化复杂，且难以观察和识别，因此织物缺陷的智能检测一直是困扰该行业的技术瓶颈多年的问题。\n\n本数据集涵盖了纺织行业中各类重要的织物缺陷，每张图片都包含一个或多个缺陷。数据包括两种类型的平纹布和花式布。其中，约8000张平纹布数据用于初赛匹配，约12000张花式布数据用于半决赛。\n\n\u003Cbr>\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 7）铝型材表面缺陷数据集（天池）\n\n- 🔗 下载链接：https:\u002F\u002Ftianchi.aliyun.com\u002Fcompetition\u002Fentrance\u002F231682\u002Finformation\n\n在铝型材的实际生产过程中，由于多种因素的影响，其表面会出现裂纹、剥落、划痕等缺陷，这将严重影响铝型材的质量。为确保产品质量，需要进行人工目视检查。然而，铝型材表面本身存在纹理，这些纹理与缺陷往往难以区分。\n\n传统的手工目视检查方法存在诸多不足，不仅工作量大、无法及时准确判断表面缺陷，而且质量检测效率也难以控制。近年来，深度学习在图像识别等领域取得了快速发展。铝型材生产企业迫切希望利用最新的AI技术来革新现有的质量检测流程，自动完成质检任务，减少漏检现象，从而提高产品质量。借助AI技术，尤其是深度学习，铝型材生产管理者可以彻底摆脱对产品表面质量状况无法全面掌握的困境。\n\n本次竞赛的数据集中包含了1万张来自实际生产中存在缺陷的铝型材监控图像，每张图像都包含一个或多个缺陷。用于机器学习的样本图像将明确标注图像中所含缺陷的类型。\n\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_32dbe72ffff9.png\">\u003C\u002Fdiv>\n\u003Cbr>\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 8）面向工业光学检测的弱监督学习（DAGM 2007）\n\n- 🔗 下载链接：https:\u002F\u002Fhci.iwr.uni-heidelberg.de\u002Fnode\u002F3616\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_23614825725f.png\">\u003C\u002Fdiv>\n\n\u003Cbr>\n\n数据集介绍：\n\n- 主要针对具有纹理背景上的各类缺陷。\n\n- 训练数据采用较弱的监督方式。\n\n- 包含十个数据集，前六个为训练数据集，后四个为测试数据集。\n\n- 每个数据集包含1000张“无缺陷”图像和150张“缺陷”图像，以灰度8位PNG格式保存。每个数据集由不同的纹理模型和缺陷模型生成。\n\n- “无缺陷”图像的背景纹理中没有任何缺陷，“缺陷”图像的背景纹理中恰好有一个标记的缺陷。\n\n- 所有数据集已被随机划分为大小相等的训练和测试子数据集。\n\n- 弱标签用椭圆表示，大致指示缺陷区域。\n\n\u003Cbr>\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 9）建筑表面裂缝\n\nCrackForest数据集是一个标注好的道路裂缝图像数据库，能够反映城市道路表面的整体状况。\n\n- Github链接：https:\u002F\u002Fgithub.com\u002Fcuilimeng\u002FCrackForest-dataset \n\n- 下载链接：https:\u002F\u002Fpan.baidu.com\u002Fs\u002F1108j5QbDr7T3XQvDxAzVpg（提取码：``jajn``）\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_f7151187a678.png\">\u003C\u002Fdiv>\n\n\u003Cbr>\n\n\u003Cp align=center>图2. 桥梁裂缝（左）和路面裂缝。\u003C\u002Fp>\n\n- \u003Cb>桥梁裂缝\u003C\u002Fb>。共有2688张未提供像素级真实标签的桥梁裂缝图像。作者为“Liangfu Li, Weifei Ma, Li Li, Xiaoxiao Gao”。文件可通过访问 https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ftree\u002Fmaster\u002FBridge_Crack_Image 获取。\n\n- \u003Cb>路面裂缝\u003C\u002Fb>。作者为Shi Yong、Cui Limeng、Qi Zhiquan、Meng Fan和Chen Zhensong。原始数据集可在 https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ftree\u002Fmaster\u002FCrackForest 找到。我们从中提取了具有像素级真实标签的图像文件。\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 10）磁性瓷砖数据集\n\n磁性瓷砖数据集由github用户abin24提供，可从 [https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ftree\u002Fmaster\u002FMagnetic-Tile-Defect](https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ftree\u002Fmaster\u002FMagnetic-Tile-Defect) 下载。该数据集曾用于其论文《磁性瓷砖表面缺陷显著性》，论文可通过 [这里](https:\u002F\u002Flink.springer.com\u002Farticle\u002F10.1007\u002Fs00371-018-1588-5) 或 [这里](https:\u002F\u002Fieeexplore.ieee.org\u002Fdocument\u002F8560423) 查阅。\n\n![dataset](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_543b47e40714.jpg) \n\n\u003Cp align=center>图3. 我们数据集的概览。\u003C\u002Fp>\n\n这也是论文《磁性瓷砖表面缺陷显著性》所使用的数据集。收集了6种常见的磁性瓷砖缺陷图像，并为其标注了像素级真实标签。\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 11）RSDDs：轨道表面缺陷数据集\n\nRSDDs数据集包含两类数据集：第一类是I型RSDDs数据集，拍摄于快速车道，包含67张具有挑战性的图像；第二类是II型RSDDs数据集，拍摄于普通\u002F重载运输轨道，包含128张具有挑战性的图像。\n\n这两类数据集中的每一张图像都至少包含一个缺陷，且背景复杂、噪声较多。\n\nRSDDs数据集中这些缺陷均由轨道表面检测领域的专业人工观察者标记。\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_f3b15d469e88.jpg\">\u003C\u002Fdiv>\n\u003Cbr>\n\n- 官方链接：http:\u002F\u002Ficn.bjtu.edu.cn\u002FVisint\u002Fresources\u002FRSDDs.aspx\n\n- 下载链接：https:\u002F\u002Fpan.baidu.com\u002Fshare\u002Finit?surl=svsnqL0r1kasVDNjppkEwg（提取码：``nanr``）\n\n\u003Cbr>\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 12）Kylberg纹理数据集 v.1.0 \n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Fimgconvert.csdnimg.cn\u002FaHR0cHM6Ly9tbWJpei5xcGljLmNuL21tYml6X3BuZy9aTmRoV05pYjNJUkJkeklpYVlQQTJ5ZmFXaFRMcVF1UElVdmxPTkVRYURGQzdaT3dWOxBhZWtCckNjQ2FxY0dWb2lhdHk2ZWszRlNTSXhjWVIwelI5TUZIZy82NDA?x-oss-process=image\u002Fformat,png\">\u003C\u002Fdiv>\n\u003Cp align=center>图4. 来自28个纹理类别的示例补丁。\u003C\u002Fp>\n\n简要说明：\n- 共28个纹理类别，见图4。\n- 每个类别包含160个独特的纹理补丁。（另一种数据集版本对每个原始补丁进行12次旋转，即每个类别有160×12=1920个纹理补丁）。\n- 纹理补丁尺寸：576×576像素。\n- 文件格式：无损压缩的8位PNG。\n- 所有补丁均经过归一化处理，平均值为127，标准差为40。\n- 每个纹理类别对应一个目录。\n- 文件命名格式如下：“blanket1-d-p011-r180.png”，其中“blanket1”为类别名称，“d”为原始图像样本编号（可能值为a、b、c或d），“p011”为第11个补丁，“r180”表示补丁旋转了180度。\n\n🔗 官方链接：http:\u002F\u002Fwww.cb.uu.se\u002F~gustaf\u002Ftexture\u002F\n\u003Cbr>\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 13）KTH-TIPS数据库\n\n重复背景纹理的数据集，示例如下：\n\n\u003Cdiv align=center>\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_2483718ef397.png\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_d5f956cace04.png\">\n\u003C\u002Fdiv>\n\n- 官方链接：https:\u002F\u002Fwww.nada.kth.se\u002Fcvap\u002Fdatabases\u002Fkth-tips\u002Fdownload.html\n\n- 下载链接：\n\n  - 数据集1：https:\u002F\u002Fpan.baidu.com\u002Fs\u002F173h8V66yRmtVo5rc2P7J4A\n\n  - 数据集2：https:\u002F\u002Fpan.baidu.com\u002Fs\u002F1dXFKn6v2PV5QS9m8gWlifA\n\n\u003Cbr>\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n\n### 14）自动扶梯踏板缺陷数据集 \n\n🔗 官方链接：https:\u002F\u002Faistudio.baidu.com\u002Faistudio\u002Fdatasetdetail\u002F44820\n\n\u003Cbr>\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 15）输电线路绝缘子数据集\n\n该数据集中，“Normal_Insulators”包含600张无人机拍摄的正常绝缘子图像。“Defective_Insulators”则包含缺陷绝缘子，共计248张缺陷图像。数据集包括数据和标签。\n\n\u003Cdiv align=center>\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_a13a48c416ed.png\">\u003C\u002Fdiv>\n\n🔗 官方链接：https:\u002F\u002Fgithub.com\u002FInsulatorData\u002FInsulatorDataSet\n\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 16）MVTEC ITODD\n\n**MVTec 工业 3D 物体检测数据集 (MVTec ITODD)** 是一个公开数据集，用于 3D 物体检测和位姿估计，特别针对工业场景和应用。\n\n该数据集包含：\n\n- 28 种物体和 3500 个带有这些物体实例标注的场景\n- 五个传感器（两个 3D 传感器和三个灰度相机）对每个场景进行观测\n\n更多信息请参阅[此 PDF 文件](https:\u002F\u002Fwww.mvtec.com\u002Ffileadmin\u002FRedaktion\u002Fmvtec.com\u002Fcompany\u002Fresearch\u002Fdatasets\u002Fmvtec_itodd.pdf) 🔍。\n\n\u003Cdiv align=center>\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_e2380bbc79f1.png\">\n\u003C\u002Fdiv>\n\n🔗 下载链接 https:\u002F\u002Fwww.mvtec.com\u002Fcompany\u002Fresearch\u002Fdatasets\u002Fmvtec-itodd\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 17）BSData - 实例分割与工业磨损预测数据集\n\n该数据集包含 1104 张三通道图像，并配有 394 个关于“点蚀”表面损伤类型的图像标注。这些标注使用 [labelme](https:\u002F\u002Fgithub.com\u002Fwkentaro\u002Flabelme) 标注工具完成，以 ``JSON`` 格式提供，因此可以转换为 VOC 和 COCO 格式。所有图像均来自两种 BSD 类型。\n\n另一种 BSD 类型包含 325 张图像，分为两种尺寸。由于该类型的所有图像都是连续拍摄的，因此污损程度在不断变化。\n\n此外，数据集还包含如上所述的 27 个点蚀发展序列，每个序列包含 69 张图像。\n\n\u003Cp align=center>\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_31159abccd9f.png\">\n  图 5. 左侧为图像示例，右侧为对应的 PNG 标注。\n\u003C\u002Fp>\n\n🔗 官方链接 https:\u002F\u002Fgithub.com\u002F2Obe\u002FBSData\n\n> 衷心感谢 @Beñat Gartzia 的推荐及您的关注！\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 18）齿轮检测数据集\n\n齿轮检测数据集 (GID) 是由百度（中国）公司举办的“全国人工智能创新应用大赛”所使用的竞赛数据集。它包含两千张灰度图像，共 28575 个标注，涉及三种真实来源的缺陷类型。每张图片的缺陷信息都记录在一个单独的 JSON 文件中，包括图像名称、标签类别、边界框以及用于分割的多边形。然而，标签类别并未明确说明具体缺陷类型，仅以数字表示，因此难以与其他相关数据集进行对比。\n\n\u003Cp align=center>\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_b9bfa9450a88.jpg\">\n    图 6. 验证测试图像及其标签示例。\n\u003C\u002Fp>\n\n🔗 官方链接 http:\u002F\u002Fwww.aiinnovation.com.cn\u002F#\u002FdataDetail?id=34\n\n- 下载链接：\n  - 齿轮检测训练数据集：https:\u002F\u002Fpan.baidu.com\u002Fs\u002F17HoFfBUQGeX7G0ibkPExrw （提取码：hm7k）\n  - 齿轮检测 A 组评估数据集：https:\u002F\u002Fpan.baidu.com\u002Fs\u002F157Zf7hcTM78GhXtXI5ySFQ （提取码：2R6K）\n  - 齿轮检测 B 组评估数据集：https:\u002F\u002Fpan.baidu.com\u002Fs\u002F1OjOZotqlRSvsYLA_qH2nXA （提取码：hypd）\n  \n- 镜像：\n  - 齿轮检测训练数据集：https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1CZo-Ab5BXkTjV-b1-NIFzYMjfJQMl4nG\u002Fview?usp=share_link\n  - 齿轮检测 A 组评估数据集：https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1-0sSrmhElBseeZWICu77lzTxoOiRD8yG\u002Fview?usp=share_link\n  - 齿轮检测 B 组评估数据集：暂无镜像。\n  \n 注意：该竞赛数据集不得用于商业用途。\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n\n### 19）AeBAD 航空发动机叶片异常检测\n\n下载链接：http:\u002F\u002Fsuo.nz\u002F2IU48P\n\n真实世界航空发动机叶片异常检测数据集 (AeBAD) 包含两个子数据集：单片叶片数据集 (AeBAD-S) 和叶片视频异常检测数据集 (AeBAD-V)。与现有数据集相比，AeBAD 具有两个特点：1.) 目标样本未对齐且尺度不同；2.) 测试集和训练集中的正常样本分布存在领域偏移，这种偏移主要由光照和视角的变化引起。\n\n![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_ef046a4da7cf.png)\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n### 20）BeanTech 异常检测(BTAD)\n\n下载链接：http:\u002F\u002Fsuo.nz\u002F2JEGEi\n\nBTAD（BeanTech 异常检测）数据集是一个真实的工业异常检测数据集。该数据集共包含 2830 张 3 种工业产品的实际图像。\n\n![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_readme_151bd15e2354.png)\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n\u003Cbr>\n\n## 3. 更多优质数据集资源汇总\n\n我一直在收集表面缺陷检测数据集，但仍有许多未收录的数据集。对于本仓库尚未收录的数据集，您可以访问以下网站查看。\u003Cb>同时，欢迎大家分享新的数据集，成为本仓库的贡献者。\u003C\u002Fb>\n\n![欢迎贡献](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fcontributing-👐%20Welcome-orange.svg)\n\n|来源|网址|推荐|\n|--|--|--|\n|Kaggle|https:\u002F\u002Fwww.kaggle.com\u002Fdatasets|⭐⭐⭐⭐⭐|\n|Paper With Code |https:\u002F\u002Fpaperwithcode.com\u002Fsota|⭐⭐⭐⭐⭐|\n|AWS 开放数据注册表|https:\u002F\u002Fregistry.opendata.aws|⭐⭐⭐|\n|微软研究院开放数据|https:\u002F\u002Fmsropendata.com |⭐⭐⭐|\n|Awesome-public-datasets|https:\u002F\u002Fgithub.com\u002Fawesomedata\u002Fawesome-public-datasets |⭐⭐|\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n\u003Cbr>\n\n## 4. 表面缺陷检测相关论文\n\n我收集了一些关于表面缺陷检测的文章。主要研究对象包括金属表面、LCD 屏幕、建筑物和输电线路等缺陷或异常物体。研究方法主要包括分类法、检测法、重建法和生成法。论文的电子版（PDF）已按日期命名并存放在“Paper”文件夹中。\n\n前往 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ftree\u002Fmaster\u002FPapers\">:open_file_folder: [\u003Cb>\u003Ci>论文\u003C\u002Fi>\u003C\u002Fb>]\u003C\u002Fa>。\n\n\u003Cbr>\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n## 致谢\n\n\u003Cp>现在你可以看到这个仓库了，我们应该感谢最初将上述数据集开源的人们。他们为我们的学习和研究工作提供了极大的帮助。收集这个数据集的想法最初来源于阅读“AI算法修炼营(AI_SuanFa)”的SFXiang关于表面缺陷检测的文章，这促使我整理出一个更加全面的数据集。论文的收集则来自一位名为“庆志的小徒弟”的CSDN用户。这些论文截止到11月19日，之后我还会继续完善。\u003Cstrong>希望大家踊跃\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Fblob\u002Fmaster\u002FCONTRIBUTING.md\" target=\"_blank\">贡献\u003C\u002Fa>。\u003C\u002Fstrong>\u003C\u002Fp>\n\n\u003Cp>最后，再次感谢上述数据集的开源贡献者们。\u003C\u002Fp>\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n## 下载\n- 下载ZIP文件，点击[这里](https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Farchive\u002Fmaster.zip)\n  \u003Cbr>或者在终端运行```git clone https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection.git```\u003Cbr>\n- 中国大陆 - 百度网盘下载链接 https:\u002F\u002Fpan.baidu.com\u002Fs\u002F122WY8F5VKqm3qMirqebRQw ``提取码:i20n``\n\n👆 [\u003Cb>返回目录\u003C\u002Fb> -->](#table-of-contents)\n\n## 通知\n\n本项目最初是由许多优秀人士基于他们的论文或行业应用而贡献的。\u003Cstrong>您只能将此数据集用于研究目的。\u003C\u002Fstrong>\n\n如果您有任何问题或想法，请随时与我联系：email: yidazhang1@gmail.com\n\n## 🍮 社区\n- Github \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Fdiscussions\" target=\"_blank\">讨论区 💬\u003C\u002Fa> 或 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Fissues\" target=\"_blank\">问题反馈 💭\u003C\u002Fa>\n\n- QQ群：734758251（密码：哈哈哈）\n- 微信群号：Yida_Zhang2\n- 邮箱：yidazhang1#gmail.com \n\n\u003Cbr>\n\n\u003Ca href=\"https:\u002F\u002Fcharmve.github.io\u002Fsponsor.html\">\u003Cimg align=\"left\" alt=\"支持我们！\" src=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002Fcomputer-vision-in-action\u002Fblob\u002Fmain\u002Fres\u002Fui\u002Ffrontpage\u002F2020-sponsors.svg\" height=\"120\" title=\"做你喜欢的事，并做到最好！\"\u002F>\u003C\u002Fa>\n\n##    支持\n     通过成为赞助商来支持这个项目。您的姓名和\u002F或 logo 将会出现在\u003Cb>我们的主页\u003C\u002Fb>上，并附带指向您网站的链接。🙏\n\n\u003Cp align=\"center\">\n    \u003Ca href=\"https:\u002F\u002Fcharmve.github.io\u002Fsponsor.html\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fcdn.buymeacoffee.com\u002Fbuttons\u002Fv2\u002Fdefault-red.png\" alt=\"赞助该项目\" width=\"280\" >\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003C!-- 后援者\n\n\u003Cdiv align=\"left\">\n    \u003Ca href=\"https:\u002F\u002Fopencollective.com\u002Fsurfacedd#sponsors\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fopencollective.com\u002Fl0cv\u002Fsponsors.svg?width=200\">\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n\n\u003Cbr>\n\u003Cdiv class=\"sponsor-level\">\n    \u003Ch2 align=\"center\">我们的后援者\u003C\u002Fh2>\n    \u003Ch4 align=\"center\">\u003Cem>期待你的加入，用代码传递爱心！\u003C\u002Fem>\u003C\u002Fh4>\n    \u003C!-- \u003Ch3>\u003Cem>期待你的加入，用代码传递爱心！\u003C\u002Fem>\u003C\u002Fh3> --\n    \u003Ca href=\"https:\u002F\u002Fopencollective.com\u002Fsurfacedd#backers\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fopencollective.com\u002Fsurfacedd\u002Fbackers.svg?width=200\">\u003C\u002Fa>\n    \u003Ctable align=\"center\">\n      \u003Ctbody>\n\t \u003Cth>\n\t   \u003Ctd>\n\t\t   日期 📅 \n\t   \u003C\u002Ftd>\n\t   \u003Ctd>\n\t\t   赞助人 👵 \n\t   \u003C\u002Ftd>\n\t   \u003Ctd>\n\t\t   资金 💰\n\t   \u003C\u002Ftd>\n\t   \u003Ctd>\n\t\t   支付方式 ♥️\n\t   \u003C\u002Ftd>\n\t   \u003Ctd>\n\t\t   留言 :speech_balloon:\n\t   \u003C\u002Ftd>\n\t \u003C\u002Fth>\n\t \u003Ctr>\n           \u003Ctd>\n\t\t   01\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   2022.02.06\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   @大巧不工\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   ￥6.66 \n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   微信红包\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   留言\n\t   \u003C\u002Ftd>\n         \u003C\u002Ftr>\n\t \u003Ctr>\n           \u003Ctd>\n\t\t   02\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   2022.02.08\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   @大巧不工\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   ￥6\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   微信红包\n\t   \u003C\u002Ftd>\n           \u003Ctd>\n\t\t   留言\n\t   \u003C\u002Ftd>\n         \u003C\u002Ftr>\n      \u003C\u002Ftbody>\n    \u003C\u002Ftable>\n    \u003Cbr>\n    \u003Csup>* \u003Cb>注\u003C\u002Fb>：由于微信和支付宝不支持个人接口调用，赞助名单上的信息是我手动添加的，更新可能会有延迟。如有需要，您也可以提交 Issue，说明您的赞助情况以及希望获得的帮助。我会在添加或解决问题后关闭相应的 Issue。\u003C\u002Fsup>\n\u003C\u002Fdiv>\n\n--->\n\n## 引用\n使用以下 bibtex 格式引用本仓库：\n```\n@misc{Surface Defect Detection,\n  title={Surface Defect Detection: Dataset and Papers},\n  author={Charmve},\n  year={2020.09},\n  publisher={Github},\n  journal={GitHub repository},\n  howpublished={\\url{https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection}},\n}\n```\n\n## 星标数量随时间变化\n\n[![星标数量随时间变化](https:\u002F\u002Fstarchart.cc\u002FCharmve\u002FSurface-Defect-Detection.svg)](https:\u002F\u002Fstarchart.cc\u002FCharmve\u002FSurface-Defect-Detection)\n\n\u003Cbr>\n\u003Cp align=\"center\">欢迎提出任何问题，如果你觉得有什么可以改进的地方，也请随时发起 PR！\u003C\u002Fp>\n\u003Ch2 align=\"center\">🌟请给这个仓库加星🌟\u003C\u002Fh2>\n\u003Cp align=\"center\">由 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\">Charmve\u003C\u002Fa> 和 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FMaiweiAI\">maiwei.ai\u003C\u002Fa> 社区创建 | 部署于 \u003Ca href=\"https:\u002F\u002Fwww.kaggle.com\u002Fyidazhang07\u002Fbridge-cracks-image\">Kaggle\u003C\u002Fa>\u003C\u002Fp>\n\n\u003Cbr>\n* \u003Ci>更新于 2021年9月17日 @\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCharmve\" target=\"_blank\">Charmve\u003C\u002Fa>, \n    \u003Ca class=\"github-button\"\n        href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\"\n        data-icon=\"octicon-star\" data-show-count=\"true\"\n        aria-label=\"在 GitHub 上为 Charmve\u002FSurface-Defect-Detection 加星\">加星\u003C\u002Fa> \n    和 \n    \u003Ca class=\"github-button\"\n        href=\"https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection\u002Ffork\"\n        data-icon=\"octicon-repo-forked\" data-show-count=\"true\"\n        aria-label=\"在 GitHub 上为 Charmve\u002FSurface-Defect-Detection 分叉\">分叉\u003C\u002Fa>\n\u003C\u002Fi>","# Surface-Defect-Detection 快速上手指南\n\n本项目是一个专注于工业表面缺陷检测的开源资源库，汇集了该领域重要的数据集、关键论文及解决方案。本指南将帮助你快速获取数据并了解核心概念。\n\n## 环境准备\n\n本项目主要提供数据集索引和学术资源，不涉及单一的可执行二进制文件安装。使用前请确保你的开发环境满足以下要求：\n\n*   **操作系统**：Linux (推荐 Ubuntu 18.04\u002F20.04), macOS 或 Windows\n*   **编程语言**：Python 3.6+\n*   **深度学习框架**：PyTorch 或 TensorFlow (根据你选择复现的具体论文代码而定)\n*   **依赖库**：\n    ```bash\n    pip install opencv-python numpy matplotlib pandas tqdm\n    ```\n*   **硬件要求**：若需训练深度学习模型，建议配备 NVIDIA GPU (支持 CUDA)；若仅用于数据浏览或传统算法验证，CPU 即可。\n\n## 安装与数据获取\n\n由于本项目核心价值在于数据集，\"安装\"过程主要为克隆仓库和下载数据。\n\n### 1. 克隆项目仓库\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FCharmve\u002FSurface-Defect-Detection.git\ncd Surface-Defect-Detection\n```\n\n### 2. 下载数据集\n项目整理了多个工业缺陷数据集（如 NEU-CLS, elpv-dataset, DeepPCB 等）。推荐使用国内镜像源以提高下载速度。\n\n**方式一：百度网盘（推荐国内用户）**\n*   **链接**: https:\u002F\u002Fpan.baidu.com\u002Fs\u002F1GWQ_acTF5BnJgpJRSw8BKA\n*   **提取码**: `o7p5`\n*   **操作**: 下载后解压至项目根目录或自定义数据目录。\n\n**方式二：Google Drive**\n*   **链接**: https:\u002F\u002Fdrive.google.com\u002Fdrive\u002Ffolders\u002F1q7lirc_yQBXxUSECwX1UvV1TS4eioFm8\n\n**注意**：部分数据集（如 NEU-CLS）的官方原始链接可能失效，请优先使用上述整理后的备份链接。具体每个数据集的详细说明请参阅项目内的 `Papers` 文件夹或对应数据集章节。\n\n## 基本使用\n\n本项目作为资源汇总，使用方式取决于你具体想要研究的数据集或复现的算法。以下是加载典型数据集（以 NEU-CLS 钢铁表面缺陷为例）进行初步查看的 Python 示例。\n\n假设你已下载并将 `NEU-CLS` 数据集放置在 `.\u002Fdata\u002FNEU-CLS` 目录下：\n\n```python\nimport os\nimport cv2\nimport matplotlib.pyplot as plt\n\n# 配置数据路径\ndata_dir = '.\u002Fdata\u002FNEU-CLS'\nclass_names = ['RS', 'Pa', 'Cr', 'PS', 'In', 'Sc'] # 六种缺陷类型\n\ndef show_sample_defect(class_idx):\n    \"\"\"显示指定类别的样本图片\"\"\"\n    class_name = class_names[class_idx]\n    class_path = os.path.join(data_dir, class_name)\n    \n    if not os.path.exists(class_path):\n        print(f\"未找到目录：{class_path}, 请确认数据已下载\")\n        return\n\n    # 获取该类下的第一张图片\n    img_files = [f for f in os.listdir(class_path) if f.endswith('.jpg') or f.endswith('.png')]\n    if not img_files:\n        print(f\"类别 {class_name} 下无图片\")\n        return\n        \n    img_path = os.path.join(class_path, img_files[0])\n    img = cv2.imread(img_path)\n    img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n\n    # 展示图片\n    plt.figure(figsize=(8, 8))\n    plt.imshow(img_rgb)\n    plt.title(f'Defect Type: {class_name}')\n    plt.axis('off')\n    plt.show()\n\n# 示例：查看第一种缺陷 (Rolling Scale)\nshow_sample_defect(0)\n```\n\n### 下一步建议\n1.  **查阅论文**：进入 `Papers` 目录，阅读 2017 年以来的关键论文，理解针对小样本（Small Sample）和实时性（Real-time）问题的解决方案。\n2.  **选择模型**：根据具体任务（分类、定位、分割），在 GitHub 上搜索对应数据集名称（如 \"NEU-CLS PyTorch\"）寻找具体的训练代码实现。\n3.  **数据增强**：针对工业场景样本少的问题，参考文档中提到的镜像、旋转、合成等数据增强策略预处理数据。","某汽车零部件质检团队正致力于升级产线，利用机器视觉自动识别金属冲压件表面的划痕、凹坑等微小缺陷。\n\n### 没有 Surface-Defect-Detection 时\n- **数据收集困难**：团队需耗费数周时间在全球各大网站零散搜索工业缺陷数据集，且难以验证数据的权威性与标注质量。\n- **算法选型迷茫**：面对层出不穷的深度学习论文，工程师缺乏系统性的综述指引，难以快速定位适合小样本或实时检测的关键算法。\n- **冷启动成本高**：从零复现经典模型（如针对钢材表面的 NEU-CLS 基准测试）需要大量调试时间，导致项目初期进度严重滞后。\n- **场景适配性差**：自行设计的成像方案往往因缺乏行业参考，导致光照不均或缺陷特征不明显，模型训练效果大打折扣。\n\n### 使用 Surface-Defect-Detection 后\n- **一站式获取资源**：直接下载该项目整合的 Google Drive\u002F百度云高质量开源数据集（如 Severstal 钢铁缺陷库），立即启动模型训练。\n- **精准锁定前沿技术**：通过项目整理的 2017 年至今的核心论文集，快速掌握解决“小样本”和“实时性”痛点的 SOTA 方案。\n- **加速基线构建**：基于项目中提供的成熟分类、定位及分割任务基准，团队在两天内便完成了首个高精度验证模型的部署。\n- **优化成像策略**：参考项目中关于不同材质表面成像特性的总结，迅速调整产线光源布局，显著提升了缺陷图像的对比度与清晰度。\n\nSurface-Defect-Detection 通过聚合全球顶尖的数据与学术成果，将工业缺陷检测项目的研发周期从数月缩短至数周，成为连接理论研究与落地应用的高效桥梁。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCharmve_Surface-Defect-Detection_543b47e4.jpg","Charmve","Wei ZHANG","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002FCharmve_3126157e.jpg","Founder of @MaiweiAI Lab, @UFund-Me  and @DeepVTuber. My research interests lie at AI Infra, Machine Learning and Computer Vision.","公众号: 迈微AI研习社","Suzhou, Beijing, Shanghai, Hongkong","yidazhang1@gmail.com",null,"charmve.github.io","https:\u002F\u002Fgithub.com\u002FCharmve",[83,87,91,95],{"name":84,"color":85,"percentage":86},"Python","#3572A5",59.3,{"name":88,"color":89,"percentage":90},"C++","#f34b7d",38.5,{"name":92,"color":93,"percentage":94},"C","#555555",1.4,{"name":96,"color":79,"percentage":97},"QMake",0.9,3986,598,"2026-04-04T15:32:19","MIT","","未说明",{"notes":105,"python":103,"dependencies":106},"该项目主要是一个表面缺陷检测的数据集和论文综述仓库，README 中未提供具体的代码运行环境、依赖库或安装指南。文中提到了深度学习中的小样本问题和实时性问题，并列举了多个数据集（如 NEU-CLS, elpv-dataset 等）的下载链接，但未包含可执行模型的特定硬件或软件版本要求。",[],[15,16,14],[109,110,111,112,113,114,115,116,117,118,119],"surface-detection","surface-defects","image-segmentation","pcb-surface-defect","surface-defect-detection","paper","defects","dataset","surface","deep-learning","charmve","2026-03-27T02:49:30.150509","2026-04-07T13:28:55.095431",[],[124],{"id":125,"version":126,"summary_zh":79,"released_at":127},134938,"v1.0","2020-12-11T07:55:15"]