---
title: KolektorSDD2
canonical_url: "https://www.modelscope.cn/datasets/OpenDataLab/KolektorSDD2"
md_url: "https://www.modelscope.cn/datasets/OpenDataLab/KolektorSDD2.md"
repository: OpenDataLab/KolektorSDD2
last_updated: 2024-07-08
license: "[CC BY-NC-SA 4.0]"
storage_size: "879 MB"
domain:
  - publishDate
  - paperUrl
  - publishUrl
  - mediaTypes
  - displayName
  - taskTypes
  - labelTypes
  - publisher
tasks:
  - 2021
  - "https://arxiv.org/pdf/2104.06064v3.pdf"
  - "https://www.vicos.si/resources/kolektorsdd2/"
  - Image
  - "KolektorSDD2 (Kolektor Surface-Defect Dataset 2)"
  - "Semantic Segmentation"
  - Semantic_seg_map
  - "University of Ljubljana"
downloads: 23933
stars: 0
---

# KolektorSDD2

> KolektorSDD2 - OpenDataLab 在 ModelScope 开源的数据集。displayName: KolektorSDD2 (Kolektor Surface-Defect Dataset 2) labelTypes: Semanticsegmap license: CC BY-NC-SA 4.0 mediaTypes: Image paperUrl: https://arxiv.org/pdf/2104.06064v3.pdf publishDate: "2021"…

OpenDataLab/KolektorSDD2 是 ModelScope 魔搭社区上的2021、https://arxiv.org/pdf/2104.06064v3.pdf、https://www.vicos.si/resources/kolektorsdd2/数据集，涉及 publishDate、paperUrl、publishUrl 领域，存储大小 879 MB，采用 [CC BY-NC-SA 4.0] 许可。

- **Repository**: OpenDataLab/KolektorSDD2
- **License**: [CC BY-NC-SA 4.0]
- **Tasks**: 2021, https://arxiv.org/pdf/2104.06064v3.pdf, https://www.vicos.si/resources/kolektorsdd2/, Image, KolektorSDD2 (Kolektor Surface-Defect Dataset 2), Semantic Segmentation, Semantic_seg_map, University of Ljubljana
- **Domain**: publishDate, paperUrl, publishUrl, mediaTypes, displayName, taskTypes, labelTypes, publisher
- **Storage size**: 879 MB
- **Downloads**: 23933
- **Stars**: 0
- **Last updated**: 2024-07-08

Source: https://www.modelscope.cn/datasets/OpenDataLab/KolektorSDD2

---

displayName: KolektorSDD2 (Kolektor Surface-Defect Dataset 2)
labelTypes:
- Semantic_seg_map
license:
- CC BY-NC-SA 4.0
mediaTypes:
- Image
paperUrl: https://arxiv.org/pdf/2104.06064v3.pdf
publishDate: "2021"
publishUrl: https://www.vicos.si/resources/kolektorsdd2/
publisher:
- University of Ljubljana
tags:
- Image
taskTypes:
- Semantic Segmentation

---
# 数据集介绍
  ## 简介
  KolektorSDD2 是一个表面缺陷检测数据集，包含 3000 多张包含多种缺陷类型的图像，是在解决实际工业问题时获得的。数据集包括： 356 张有可见缺陷的图像 2979 张没有任何缺陷的图像 图像大小约为 230 x 630 像素 训练集有 246 个正图像和 2085 个负图像 测试集有 110 个正图像和 894 个负图像 几种不同类型的缺陷（划痕、轻微斑点、表面缺陷等）
  ## 引文
  ```
 @article{bovzivc2021mixed,
 title={Mixed supervision for surface-defect detection: From weakly to fully supervised learning},
 author={Bo{\v{z}}i{\v{c}}, Jakob and Tabernik, Domen and Sko{\v{c}}aj, Danijel},
 journal={Computers in Industry},
 volume={129},
 pages={103459},
 year={2021},
 publisher={Elsevier}
 }
 ```
  
## Download dataset
:modelscope-code[]{type="git"}