---
license: cc-by-4.0
task_categories:
  - object-detection
language:
  - nl
tags:
  - license-plates
  - dutch
  - netherlands
  - vehicles
  - trucks
  - computer-vision
  - yolo
  - coco
pretty_name: Dutch License Plates (Real-World)
size_categories:
  - 1K<n<10K
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*
---

# Dutch License Plates — Real-World Dataset

Real-world photos of Dutch vehicles with bounding-box annotations for license
plates. 

Splits: **70 % train / 15 % validation / 15 % test** (reproducible seed=42).

## Usage

```python
from datasets import load_dataset

ds = load_dataset("Rickkosse/license-plates")

```

## Bounding-box format

Annotations follow **COCO format**: `[x, y, width, height]` in pixels
(top-left corner + dimensions, not normalised).

## Fields

| Field | Type | Description |
|-------|------|-------------|
| `image` | Image | JPEG photo |
| `image_id` | int | Original COCO image id |
| `width` / `height` | int | Image dimensions |
| `objects.bbox` | [[float×4]] | COCO bboxes `[x,y,w,h]` |
| `objects.plate_text` | [string] | Recognised plate text |
| `objects.confidence` | [float] | YOLOv11 detection confidence |
| `primary_plate` | string | Best detected plate text |
| `source` | string | Origin: `truckfan`, `platesmania` |
| `source_url` | string | Source page URL |


## Related

- Model: [Rickkosse/rfdetr_licences_plate_detector](https://huggingface.co/Rickkosse/rfdetr_licences_plate_detector)

## License & attribution

Images sourced from truckfan.nl, platesmania.com.
Annotations auto-generated with
[morsetechlab/yolov11](https://huggingface.co/morsetechlab/yolov11-license-plate-detection)
+ [fast-plate-ocr](https://github.com/ankandrew/fast-plate-ocr).