RescueAI: Real-Time Detection of Stranded Victims and Flood Severity Using Deep Learning

This paper presents an innovative approach to detect humans stranded in floodwaters and assess flood severity using deep learning techniques [1], specifically the YOLOv5 algorithm. By leveraging a dataset of flood images captured by drones, the system identifies stranded individuals and estimates the flood severity based on varying water levels. The paper integrates YOLOv5 with an ARIMA regression model to predict severity levels, using water levels from 1 to 11. A series of modules were designed to upload, preprocess, and process datasets, generate the YOLOv5 model, and provide real-time predictions for human detection from images and videos. This solution aims to enhance flood rescue operations [1] and provide crucial information about flood severity for effective disaster management. The performance of the YOLOv5 model is visualized through training graphs, showcasing its accuracy and efficiency in detecting flood victims and predicting water levels. The system is capable of running real-time analysis, making it a valuable tool for first responders in flood-affected regions.

• Publication year

  2025

• Venue

  2025 IEEE First International Conference on Innovations in Engineering and Next-Generation Technologies for Sustainability (ICINVENTS)

• Publication date

  2025-11-07

• Fields of study

  Not labeled

• Identifiers
  DOI 10.1109/ICINVENTS64613.2025.11401873
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Modeling urban coastal flood severity from crowd-sourced flood reports using Poisson regression and Random Forest

  2018influential reference
• Prediction of the future flood severity in plain river network region based on numerical model: A case study

  2017cited by this paper
• Flash Flood Detection in Urban Cities Using Ultrasonic and Infrared Sensors

  2016cited by this paper
• Exploiting Image-trained CNN Architectures for Unconstrained Video Classification

  2015cited by this paper
• C3D: Generic Features for Video Analysis

  2014cited by this paper
• A novel approach to urban flood monitoring using computer vision

  2014influential reference
• Ensemble hydro-meteorological simulation for flash flood early detection in southern Switzerland

  2012cited by this paper
• Flash flood detection through a multi-stage probabilistic warning system for heavy precipitation events

  2011cited by this paper
• Flash flood forecasting, warning and risk management: the HYDRATE project

  2011cited by this paper
• A review of advances in flash flood forecasting

  2011influential reference
• A spatially distributed flash flood forecasting model

  2008cited by this paper
• Design of early warning flood detection systems for developing countries

  2007influential reference

• No citing papers are available for this paper.