VibES / README.md
viciopoli's picture
Update README.md
12fc8a3 verified
metadata
license: apache-2.0
language:
  - en
size_categories:
  - n<1K
tags:
  - event-camera
  - neuromorphic-vision
  - motion-compensation
  - frequency-estimation
  - vibration
  - hdf5
pretty_name: VibES Event Camera Datasets
task_categories:
  - robotics
arxiv: 2508.19094
configs:
  - config_name: amiev
    data_files: amiev_*.hdf5
  - config_name: checkerpattern
    data_files: checkerpattern_*.hdf5
  - config_name: logo
    data_files: logo_*.hdf5
  - config_name: stagger
    data_files: stagger_*.hdf5
  - config_name: pattern
    data_files: pattern_*.hdf5

Dataset Card for VibES Event Camera Datasets

Dataset Description

Dataset Summary

The VibES dataset contains event-based camera recordings collected under static and mechanically induced vibration conditions. It is intended for research on motion compensation, frequency and phase estimation, and vibration-aware event-based perception.

The dataset includes both real and synthetic event streams, stored in Prophesee / Metavision HDF5 format.

Dataset Structure

Data Instances

Each instance is an HDF5 file containing asynchronous event streams captured from an event camera or generated synthetically.

Data Fields

All files follow the Prophesee / Metavision HDF5 format:

  • /CD/events: event stream (x, y, t, p)
  • /CD/indexes: event indexing
  • /EXT_TRIGGER/events: external trigger events
  • /EXT_TRIGGER/indexes: trigger indexing

Dataset Creation

Curation Rationale

The dataset was created to study how induced mechanical vibrations can improve robustness and persistence in event-based sensing.

Source Data

Initial Data Collection and Normalization

  • amiev_*: real event data from Microsaccade-inspired event camera for robotics, converted to Prophesee-style HDF5
  • checkerpattern, logo, pattern: real event-based recordings
  • stagger_*: synthetic event sequences with controlled motion
  • stagger_30hz_*_vib: synthetic vibration at approximately 30 Hz

Annotations

No manual annotations are provided.

Personal and Sensitive Information

The dataset does not contain personal or sensitive information.


Additional Information

Licensing Information

This dataset is released under the Apache License, Version 2.0.

Citation Information

If you use this dataset, please cite:

@inproceedings{polizzi_2026_3DV,
  title={VibES: Induced Vibration for Persistent Event-Based Sensing},
  author={Polizzi, Vincenzo and Yang, Stephen and Clark, Quentin and
          Kelly, Jonathan and Gilitschenski, Igor and Lindell, David B.},
  booktitle={International Conference on 3D Vision (3DV)},
  year={2026}
}

Contributions

Contributions and issues are welcome via the repository.