Datasets:
Tasks:
Image Segmentation
Modalities:
Image
Formats:
imagefolder
Languages:
English
Size:
< 1K
ArXiv:
File size: 6,150 Bytes
2e9cd44 a4a3f92 2e9cd44 672c51a 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 2e9cd44 a4a3f92 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 | ---
annotations_creators: []
language: en
size_categories:
- n<1K
task_categories:
- image-segmentation
task_ids: []
pretty_name: kolektorsdd
tags:
- fiftyone
- image
- image-segmentation
dataset_summary: '
This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 399 samples.
## Installation
If you haven''t already, install FiftyOne:
```bash
pip install -U fiftyone
```
## Usage
```python
import fiftyone as fo
from fiftyone.utils.huggingface import load_from_hub
# Load the dataset
# Note: other available arguments include ''max_samples'', etc
dataset = load_from_hub("Voxel51/Kolektor_Surface_Defect")
# Launch the App
session = fo.launch_app(dataset)
```
'
---
# Dataset Card for Kolektor Surface-Defect Dataset

KolektorSDD (Kolektor Surface-Defect Dataset) is a grayscale industrial surface-inspection
dataset of electrical commutators.
This FiftyOne dataset uses the **box-annotation release** intended for the ICPR 2021 and
COMIND 2021 papers ([download](https://go.vicos.si/kolektorsddboxes)): one sample per
surface image, with defect regions annotated as axis-aligned bounding boxes stored as filled
rectangles in the label masks.
This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 399 samples.
## Installation
If you haven't already, install FiftyOne:
```bash
pip install -U fiftyone
```
## Usage
```python
import fiftyone as fo
from fiftyone.utils.huggingface import load_from_hub
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset = load_from_hub("Voxel51/Kolektor_Surface_Defect")
# Launch the App
session = fo.launch_app(dataset)
```
## Dataset Details
- **Curated by:** Domen Tabernik, Samo Šela, Jure Skvarč, Danijel Skočaj (University of Ljubljana / ViCoS Lab); images provided and annotated by Kolektor Group d.o.o.
- **Paper (dataset):** [Segmentation-Based Deep-Learning Approach for Surface-Defect Detection](https://doi.org/10.1007/s10845-019-01476-x)
- **Box annotations used in:** [End-to-end training of a two-stage neural network for defect detection](https://arxiv.org/abs/2007.07676) (ICPR 2020) and [Mixed supervision for surface-defect detection](http://prints.vicos.si/publications/385) (Computers in Industry, 2021)
- **Project page:** https://www.vicos.si/resources/kolektorsdd/
- **Download (this release):** https://go.vicos.si/kolektorsddboxes
- **License:** [CC BY-NC-SA 4.0](http://creativecommons.org/licenses/by-nc-sa/4.0/) (non-commercial; contact [Danijel Skočaj](https://www.vicos.si/people/danijel_skocaj/) for commercial use)
### What the data contains
Images were captured in a controlled industrial environment. Each sample is one
non-overlapping view of a commutator surface. Defects are microscopic fractures or
cracks in the plastic embedding.
| Property | Value |
|---|---|
| Total images | 399 |
| Physical items (boards) | 50 (`kos01`–`kos50`) |
| Surfaces per item | 8 (`Part0`–`Part7`) |
| Defective images | 52 |
| Non-defective images | 347 |
| Image type | Grayscale JPG |
| Original size | 500 px wide × 1240–1270 px tall |
| Recommended eval size | 512 × 1408 px (per dataset authors) |
Defect visibility: for 48 items the defect appears in exactly one image; for 2 items
it appears in two images.
A separate **fine pixel-annotation release** exists for the JIM2019 paper
([download](https://go.vicos.si/kolektorsdd)). That version is not what this card describes.
### Raw download layout
```
kolektorsdd/
kos01/
Part0.jpg
Part0_label.bmp
Part1.jpg
Part1_label.bmp
...
kos02/
...
```
- `Part*.jpg` — surface image
- `Part*_label.bmp` — defect annotation mask (non-zero = defect region)
In this box-annotation release, each defective mask is a **filled axis-aligned bounding box**
around the defect, not a precise pixel-wise segmentation of the crack shape.
### Train/test splits
The authors evaluate with **3-fold cross-validation**, keeping all 8 images of the
same physical item in the same fold. Official split files:
[KolektorSDD-training-splits.zip](https://data.vicos.si/datasets/KSDD/KolektorSDD-training-splits.zip).
This FiftyOne dataset does **not** assign fold/split labels. Add them externally if needed.
---
## FiftyOne Dataset Structure
| Property | Value |
|---|---|
| Hub dataset | `harpreetsahota/Kolektor_Surface_Defect` |
| Local dataset name | `kolektorsdd` |
| Media type | `image` |
| Samples | 399 |
### Sample fields
| Field | Type | Description |
|---|---|---|
| `filepath` | `StringField` | Path to source `Part*.jpg` |
| `board_id` | `StringField` | Board directory name, e.g. `"kos01"` |
| `has_defect` | `BooleanField` | `True` if the mask contains any foreground pixel |
| `ground_truth` | `EmbeddedDocumentField(Segmentation)` | Binarized mask (`0` = background, `1` = defect) |
The local parser (`parse_to_fo.py`) reads each BMP label and stores a `{0, 1}` mask on
the sample. For defective images in this release, the foreground region is a filled
bounding box rather than a tight defect outline.
---
## Citation
**BibTeX (dataset):**
```bibtex
@article{Tabernik2019JIM,
author = {Tabernik, Domen and {\v{S}}ela, Samo and Skvar{\v{c}}, Jure and Sko{\v{c}}aj, Danijel},
journal = {Journal of Intelligent Manufacturing},
title = {{Segmentation-Based Deep-Learning Approach for Surface-Defect Detection}},
year = {2019},
month = {May},
day = {15},
issn = {1572-8145},
doi = {10.1007/s10845-019-01476-x}
}
```
**BibTeX (box annotations / mixed supervision):**
```bibtex
@article{Bozic2021COMIND,
author = {Bo{\v{z}}i{\v{c}}, Jakob and Tabernik, Domen and Sko{\v{c}}aj, Danijel},
journal = {Computers in Industry},
title = {{Mixed supervision for surface-defect detection: from weakly to fully supervised learning}},
year = {2021}
}
```
**APA:**
Tabernik, D., Šela, S., Skvarč, J., & Skočaj, D. (2019). Segmentation-Based Deep-Learning Approach for Surface-Defect Detection. *Journal of Intelligent Manufacturing*. https://doi.org/10.1007/s10845-019-01476-x
|