Twitter Flood Images with Impact Relevance Annotations

Assembled from Twitter images filtered by flood-related keywords and annotated for relevance to flood impact analysis, published by cvjena in 2020. It provides a labeled set of social media visuals to distinguish between general flood mentions and actionable impact data.

Use Cases

• Training image classification models to identify flood impact relevance labels
• Filtering social media feeds for disaster response using visual relevance annotations
• Benchmarking computer vision algorithms on noisy, user-generated content

Strengths

• Annotated specifically for flood impact relevance
• Derived from real-world social media streams
• Created by the Computer Vision Group Jena (cvjena)

Limitations

• Unknown sample size
• Potential for link rot if images are not hosted directly
• Lack of geographic or temporal metadata for the source tweets

Provenance

Source

cvjena (Computer Vision Group Jena)

Collection Method

Twitter keyword scraping followed by manual annotation

Freshness

Last updated in November 2020; likely a static release.