Cross-attention Guided CNN-Transformer-GNN Architecture for Multi-class Disaster Image Classification

Disaster image classification emerged as an important research area for supporting rapid disaster assessment, situational awareness, and emergency response operations.Nevertheless, accurately distinguishing among different disaster categories remains challenging due to large intra-class variations, complex backgrounds, and the limited capability of existing approaches to simultaneously capture local visual patterns, global contextual dependencies, and inter-feature relationships.To tackle these challenges, this study proposes a hybrid Convolutional Neural Network (CNN)-Transformer-Graph Neural Network (GNN) approach incorporating a Cross-Attention Gating Block (CAGB) for multi-class disaster image classification.The proposed architecture combines InceptionV3-based convolutional feature extraction with Transformer-based contextual representation learning and Lightweight GNN-based relational feature modeling to generate comprehensive feature representations.Furthermore, the CAGB adaptively refines and integrates heterogeneous features through cross-attention-guided gating, enabling enhanced discrimination among disaster categories.The proposed framework was assessed on a disaster image dataset comprising Cyclone, Earthquake, Flood, and Wildfire classes with 4,428 images.Experimental results demonstrated classification performance, attaining an accuracy of 96.16%, macro-average precision of 96.33%, macro-average recall of 96.10%, macro-average F1-score of 96.20%, and micro-average ROC-AUC of 99.71%.Comparative analysis with existing disaster image classification methods further verified the effectiveness of the proposed approach, highlighting its potential for intelligent disaster monitoring and decision-support applications.

Authors

Institutions

Publication Details

Journal
International journal of intelligent engineering and systems
Published
2026-11-17
DOI
https://doi.org/10.22266/ijies2026.1231.40
Primary Topic
Advanced Neural Network Applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Cross-attention Guided CNN-Transformer-GNN Architecture for Multi-class Disaster Image Classification

S. Suprakash, Rekha M. Pillai
1 links
International journal of intelligent engineering and systems
Advanced Neural Network Applications
article

Cross-attention Guided CNN-Transformer-GNN Architecture for Multi-class Disaster Image Classification

S. Suprakash, Rekha M. Pillai
article en

Abstract

Disaster image classification emerged as an important research area for supporting rapid disaster assessment, situational awareness, and emergency response operations.Nevertheless, accurately distinguishing among different disaster categories remains challenging due to large intra-class variations, complex backgrounds, and the limited capability of existing approaches to simultaneously capture local visual patterns, global contextual dependencies, and inter-feature relationships.To tackle these challenges, this study proposes a hybrid Convolutional Neural Network (CNN)-Transformer-Graph Neural Network (GNN) approach incorporating a Cross-Attention Gating Block (CAGB) for multi-class disaster image classification.The proposed architecture combines InceptionV3-based convolutional feature extraction with Transformer-based contextual representation learning and Lightweight GNN-based relational feature modeling to generate comprehensive feature representations.Furthermore, the CAGB adaptively refines and integrates heterogeneous features through cross-attention-guided gating, enabling enhanced discrimination among disaster categories.The proposed framework was assessed on a disaster image dataset comprising Cyclone, Earthquake, Flood, and Wildfire classes with 4,428 images.Experimental results demonstrated classification performance, attaining an accuracy of 96.16%, macro-average precision of 96.33%, macro-average recall of 96.10%, macro-average F1-score of 96.20%, and micro-average ROC-AUC of 99.71%.Comparative analysis with existing disaster image classification methods further verified the effectiveness of the proposed approach, highlighting its potential for intelligent disaster monitoring and decision-support applications.

International journal of intelligent engineering and systemsVol. 19(12)
Kalasalingam Academy of Research and Education (IN)
Advanced Neural Network Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon