Multi-source Multimodal Data and Deep Learning for Disaster Response: A Systematic Review.

Area of Science:

• Computer Science
• Artificial Intelligence
• Disaster Management

Background:

• Information sharing during disasters is evolving with new technologies like social media.
• Current disaster response heavily relies on text data, underutilizing other modalities (audio, video, images).
• Deep Learning (DL) shows potential for multi-modal data analysis but is largely academic in disaster response (DR) applications.

Purpose of the Study:

• To systematically review the application of Deep Learning (DL) in disaster response (DR) tasks.
• To identify successes, challenges, and opportunities in using DL for DR.
• To provide guidance for future research in this domain.

Main Methods:

• Conducted a systematic review of 83 articles on DL for DR.
• Analyzed DL applications based on components of learning (ML aspects).
• Developed a flowchart and guidance for future research.

Main Results:

• Identified current successes and challenges in applying DL to DR tasks.
• Highlighted opportunities for leveraging multi-modal data in DR through DL.
• The review provides a structured overview of DL's role in DR.

Conclusions:

• Deep Learning (DL) offers significant potential to improve disaster response (DR) by utilizing multi-modal data.
• Further research is needed to bridge the gap between academic findings and practical DR applications.
• A structured approach and guidance are crucial for realizing the full benefits of DL in DR.