Datasets:
QCRI
/

Modalities:
Image
Text
Formats:
parquet
Languages:
English
ArXiv:
Libraries:
Datasets
Dask
License:
MEDIC / README.md
Firoj's picture
Upload README.md with huggingface_hub
f2ee67b verified
metadata
license: cc-by-nc-sa-4.0
task_categories:
  - image-classification
language:
  - en
tags:
  - Disaster
  - Crisis Informatics
pretty_name: 'MEDIC: A Multi-Task Learning Dataset for Disaster Image Classification'
size_categories:
  - 10K<n<100K
dataset_info:
  splits:
    - name: train
      num_examples: 49353
    - name: dev
      num_examples: 6157
    - name: test
      num_examples: 15688

MEDIC: A Multi-Task Learning Dataset for Disaster Image Classification

Data

The MEDIC is the largest multi-task learning disaster-related dataset, an extended version of the crisis image benchmark dataset. It consists of data from several sources, including CrisisMMD, data from AIDR, and the Damage Multimodal Dataset (DMD). The dataset contains 71,198 images.

Table of Contents

Data Format and Directories

Directories

  • data: Main directory with the following subdirectories:
    • aidr_disaster_types/: Contains images collected using AIDR system for disaster types task.
    • aidr_info/: Contains images collected using AIDR system for informativeness task.
    • ASONAM17_Damage_Image_Dataset/: Damage Assessment Dataset.
    • crisismmd/: CrisisMMD dataset.
    • multimodal-deep-learning-disaster-response-mouzannar/: Damage Multimodal Dataset (DMD).
  • MEDIC_train.tsv, MEDIC_dev.tsv, MEDIC_test.tsv: Training, development, and testing files with specific file formats.
  • LICENSE_CC_BY_NC_SA_4.0.txt: License information.
  • terms-of-use.txt: Terms and conditions.

Format

  • image_id: Corresponds to the tweet ID from Twitter or ID from the respective source.
  • event_name: Name of the event or data source.
  • image_path: Relative path of the image.
  • damage_severity: Damage severity class label.
  • informative: Informativeness class label.
  • humanitarian: Humanitarian class label.
  • disaster_types: Disaster types class label.

Disaster Response Tasks

  1. Disaster Types

    • Earthquake
    • Fire
    • Flood
    • Hurricane
    • Landslide
    • Not disaster
    • Other disaster
  2. Informativeness

    • Informative
    • Not informative
  3. Humanitarian Categories

    • Affected, injured, or dead people
    • Infrastructure and utility damage
    • Not humanitarian
    • Rescue, volunteering, or donation effort
  4. Damage Severity Assessment

    • Little or no damage
    • Mild damage
    • Severe damage

Downloads

License

The MEDIC dataset is published under CC BY-NC-SA 4.0 license, which means everyone can use this dataset for non-commercial research purpose: https://creativecommons.org/licenses/by-nc/4.0/. See LICENSE_CC_BY_NC_SA_4.0.txt

Citation

Please cite the following papers if you use this dataset in your research:

  1. Firoj Alam, Tanvirul Alam, Md. Arid Hasan, Abul Hasnat, Muhammad Imran, Ferda Ofli. MEDIC: A Multi-Task Learning Dataset for Disaster Image Classification. Neural Computing and Applications, 35(3):2609–2632, 2023. paper Arxiv
  2. Firoj Alam, Ferda Ofli, Muhammad Imran, Tanvirul Alam, Umair Qazi. Deep Learning Benchmarks and Datasets for Social Media Image Classification for Disaster Response. In 2020 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM), 2020.
  3. Firoj Alam, Ferda Ofli, and Muhammad Imran. CrisisMMD: Multimodal Twitter Datasets from Natural Disasters. In Proceedings of the 12th International AAAI Conference on Web and Social Media (ICWSM), 2018, Stanford, California, USA.
  4. Hussein Mozannar, Yara Rizk, and Mariette Awad. Damage Identification in Social Media Posts using Multimodal Deep Learning. In Proc. of ISCRAM, May 2018, pp. 529–543.
  5. Dat Tien Nguyen, Ferda Ofli, Muhammad Imran, and Prasenjit Mitra. Damage assessment from social media imagery data during disasters. In Proc. of ASONAM, pages 1–8, Aug 2017.
@article{alam2022medic,
title={{MEDIC}: A Multi-Task Learning Dataset for Disaster Image Classification},
author={Firoj Alam and Tanvirul Alam and Md. Arid Hasan and Abul Hasnat and Muhammad Imran and Ferda Ofli},
Keywords = {Multi-task Learning, Social media images, Image Classification, Natural disasters, Crisis Informatics, Deep learning, Dataset},
journal={Neural Computing and Applications},
volume={35},
issue={3},
pages={2609--2632},
year={2023},
publisher={Springer}
}

@InProceedings{crisismmd2018icwsm,
  author = {Alam, Firoj and Ofli, Ferda and Imran, Muhammad},
  title = {{CrisisMMD}: Multimodal Twitter Datasets from Natural Disasters},
  booktitle = {Proceedings of the 12th International AAAI Conference on Web and Social Media (ICWSM)},
  year = {2018},
  month = {June},
  date = {23-28},
  location = {USA}
}

@inproceedings{10.1109/ASONAM49781.2020.9381294,
author = {Alam, Firoj and Ofli, Ferda and Imran, Muhammad and Alam, Tanvirul and Qazi, Umair},
title = {Deep learning benchmarks and datasets for social media image classification for disaster response},
year = {2021},
isbn = {9781728110561},
publisher = {IEEE Press},
url = {https://doi.org/10.1109/ASONAM49781.2020.9381294},
doi = {10.1109/ASONAM49781.2020.9381294},
booktitle = {Proceedings of the 12th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining},
pages = {151–158},
numpages = {8},
keywords = {benchmarking, crisis computing, deep learning, disaster image classification, natural disasters, social media},
location = {Virtual Event, Netherlands},
series = {ASONAM '20}
}

@inproceedings{mouzannar2018damage,
  title={Damage Identification in Social Media Posts using Multimodal Deep Learning.},
  author={Mouzannar, Hussein and Rizk, Yara and Awad, Mariette},
  booktitle={ISCRAM},
  year={2018},
  organization={Rochester, NY, USA}
}

@inproceedings{nguyen2017damage,
  title={Damage assessment from social media imagery data during disasters},
  author={Nguyen, Dat T and Ofli, Ferda and Imran, Muhammad and Mitra, Prasenjit},
  booktitle={Proceedings of the 2017 IEEE/ACM international conference on advances in social networks analysis and mining 2017},
  pages={569--576},
  year={2017}
}