Skip to content

Dataset · QCRI

Manually annotated disaster tweets and photos

QCRI/CrisisMMD

Multimodal corpus (10K–100K) of posts and images regarding seven major natural disasters in 2017.

What's inside

The second version of the CrisisMMD dataset includes 10,000 to 100,000 English records consisting of tweets and attached photos. The materials were collected during seven major natural disasters in 2017, including earthquakes, hurricanes, wildfires, and floods. The data is provided with three types of manual annotations for emergency response tasks.

What you can do with it

  • Training multimodal classifiers for emergency messages
  • Automated filtering and damage assessment using photos and social media text
  • Testing emergency monitoring systems
Tasks
Image classification
Languages
English
Size
10K–100K rows
Downloads this month
21.8K
Likes
6

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("QCRI/CrisisMMD")