/apron-dataset

Information and scripts for the Apron Dataset

Primary LanguagePythonOtherNOASSERTION

The Apron Dataset

apron_dataset_overview.png

This repository provides all relevant information and utility scripts of the Apron Dataset, which focuses on training and evaluating classification and detection models for airport-apron logistics. A detailed description can be found in the corresponding publication. The image data and annotations are available on request at wilddash.cc/aprondataset.

Annotation Format

Annotations are provided as csv files for each image, defining object instances by a bounding box, label id and the following meta parameters:

  • occluded: non-occluded (0), occluded (1)
  • atmosphere: clear (0), light rain (1), heavy rain (2), fog (3), snow (4)
  • lighting: sunny (0), diffuse (1), artificial (2)
  • timeofday: day (0), twilight (1), night (2)
  • degradation: low (0), high (1)

The specifications of label ids, mappings and visualization colors are defined in datasets.py.

Dataset Variants

The provided dataset contains fine-grained annotations for 43 categories (ApronFine). Use map_dataset.py to map them to arbitrary target datasets. Mappings of the default variants ApronTop and ApronCoarse are provided.

Visualization

Use visualize_annotations.py to overlay annotations as colored bounding boxes on all images in a given directory.

Licence

The Apron Dataset is released to academic and non-academic entities for non-commercial purposes such as academic research, teaching, scientific publications or personal experimentation (LICENCE).

Citing

If you use the dataset for your research, please use the following BibTeX entry:

@InProceedings{Steininger_2022_ACCV,
    author    = {Steininger, Daniel and Kriegler, Andreas and Pointner, Wolfgang and Widhalm, Verena and Simon, Julia and Zendel, Oliver},
    title     = {Towards Scene Understanding for Autonomous Operations on Airport Aprons},
    booktitle = {Proceedings of the Asian Conference on Computer Vision (ACCV) Workshops},
    month     = {December},
    year      = {2022},
    pages     = {147-163}
}