README.md

---
license: cc-by-4.0
pretty_name: Automingo-VQA
language:
- en
task_categories:
- visual-question-answering
- image-to-text
tags:
- autonomous-driving
- adas
- euro-ncap
- vision-language
- vqa
- safety-critical
- driving
- multimodal
- benchmark
size_categories:
- 1K<n<10K

dataset_info:
  features:
  - name: scene_id
    dtype: string
  - name: situation
    dtype: string
  - name: question
    dtype: string
  - name: ground_truth_answer
    dtype: string
  - name: ground_truth_reasoning
    dtype: string
  - name: time_span
    list: float32
  - name: distractor_1
    dtype: string
  - name: distractor_2
    dtype: string
  - name: distractor_3
    dtype: string
  - name: image_1
    dtype: image
  - name: image_2
    dtype: image
  - name: image_3
    dtype: image
  - name: image_4
    dtype: image
  - name: image_5
    dtype: image
  splits:
  - name: train
    num_bytes: 16789771017
    num_examples: 3256
  - name: validation
    num_bytes: 5549085751
    num_examples: 1055
  download_size: 21677589117
  dataset_size: 22338856768
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*

  
---
# Automingo-VQA

Automingo-VQA is a vision-language dataset for safety-critical autonomous driving scene understanding. It is designed for Visual Question Answering (VQA), event-level reasoning, and structured analysis of ADAS-relevant traffic scenarios.

The dataset focuses on real-world driving situations aligned with Euro NCAP-related safety scenarios, including cut-ins, traffic light transitions, vulnerable road user interactions, braking conflicts, construction zones, intersections, roundabouts, and speed limit adaptation.

Each sample is built around a driving event or an empty/negative segment and contains temporally structured front-camera frames paired with expert-annotated questions, answers, and reasoning.

## Dataset Details

Automingo-VQA contains:

- 6,565 real-world driving images
- 1,313 events
- 771 safety-critical events
- 542 empty / negative events
- 5,792 question-answer pairs
- Five temporally sampled frames per event
- Expert-annotated answers and reasoning
- Multiple-choice validation examples
- Anonymised faces and license plates

The data were collected in and around the Madrid metropolitan area, Spain, using front-facing egocentric vehicle cameras.

## Scenarios

The dataset covers the following scenario categories:

| Abbreviation | Scenario |
|---|---|
| CI | Cut-In |
| TL | Traffic Light |
| VRU | Vulnerable Road User |
| ML | Merging Lane |
| LPC | Lateral Parked Car |
| LB | Leading Braking |
| CS | Construction Site |
| CO | Crossing Object |
| IS | Intersection |
| RA | Roundabout |
| SLA | Speed Limit Adaptation |

## Intended Uses

This dataset is intended for:

- Visual Question Answering in autonomous driving
- Vision-Language Model training and evaluation
- ADAS perception analysis
- Safety-critical traffic scene understanding
- Event-level driving reasoning
- Benchmarking multimodal models on structured driving scenarios

### Out-of-Scope Uses

This dataset should not be used for:

- Identifying people, vehicles, or license plates
- Deanonymisation attempts
- Surveillance applications
- Automated legal, insurance, or enforcement decisions
- Deployment as the sole validation source for safety-critical autonomous driving systems

## Limitations

Automingo-VQA is focused on a defined set of ADAS-relevant scenarios and should not be considered a complete representation of all possible driving situations.

Known limitations include:

- Single front-facing camera perspective
- Data collected mainly around Madrid, Spain
- Limited coverage of weather, lighting, and geographic diversity
- Potential ambiguity in complex scenarios such as roundabouts and intersections
- Possible residual annotation or distractor-generation errors

## Paper

This dataset is associated with the paper:

**Automingo: Seeing the Unseen - Vision-Language Edge Case Dataset for Detection and Analysis of Autonomous Driving**

Paper URL: [https://openaccess.thecvf.com/content/CVPR2026W/AUTOPILOT/papers/Divis_Automingo_Seeing_the_Unseen_-_Vision-Language_Edge_Case_Dataset_for_CVPRW_2026_paper.pdf](https://openaccess.thecvf.com/content/CVPR2026W/AUTOPILOT/papers/Divis_Automingo_Seeing_the_Unseen_-_Vision-Language_Edge_Case_Dataset_for_CVPRW_2026_paper.pdf)

## Authors

- Václav Diviš
- Íñigo Barceló Álvarez
- Alejandro Fariñas Nubla
- Enrique Sánchez
- Antonio Hernández-Ros Briales
- Ondřej Valach
- Ivan Gruber
- Marek Hruž

### Affiliations

- ARRK Engineering GmbH
- University of West Bohemia, Faculty of Applied Sciences, Department of Cybernetics and New Technologies for the Information Society

## License

This dataset is released under the **Creative Commons Attribution 4.0 International License (CC BY 4.0)**.

You are free to share, copy, redistribute, adapt, transform, and build upon the dataset for any purpose, including commercial use, provided that appropriate credit is given to the original authors.

License: [https://creativecommons.org/licenses/by/4.0/](https://creativecommons.org/licenses/by/4.0/)

## Citation

If you use this dataset in your research, product development, benchmark, model training, evaluation pipeline, or any derived work, please cite the associated paper:

```bibtex
@inproceedings{Divis_2026_CVPR,
  title     = {Automingo: Seeing the Unseen - Vision-Language Edge Case Dataset for Detection and Analysis of Autonomous Driving},
  author    = {Diviš, Václav and Barceló Álvarez, Íñigo and Fariñas Nubla, Alejandro and Sánchez, Enrique and Hernández-Ros Briales, Antonio and Valach, Ondřej and Gruber, Ivan and Hruž, Marek},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
  month     = {June},
  year      = {2026},
  pages     = {665-674}
}
```