Slide
Centre Interdisciplinaire
de Recherche et d’Innovation
en Cybersécurité et Société
de Recherche et d’Innovation
en Cybersécurité et Société
1.
Sadallah, N.; Allili, M. S.
Crackforward: Context-Aware Severity-Stage Crack Synthesis for Data Augmentation Proceedings Article
In: Can Conf Electr Comput Eng, pp. 627–632, Institute of Electrical and Electronics Engineers Inc., 2026, ISBN: 08407789 (ISSN); 979-833158840-3 (ISBN), (Journal Abbreviation: Can Conf Electr Comput Eng).
Abstract | Links | BibTeX | Tags: Context-Aware, Crack Expansion, Crack morphology, Crack propagation, Crack segmentation, Crack segmentations, Data augmentation, Growth patterns, Health monitoring, Local Eigenvector Orientation, Morphology, Random processes, Structural health, Structural health monitoring, Textures, Unet-style generation
@inproceedings{sadallahCrackforwardContextAwareSeverityStage2026,
title = {Crackforward: Context-Aware Severity-Stage Crack Synthesis for Data Augmentation},
author = {N. Sadallah and M. S. Allili},
url = {https://www.scopus.com/pages/publications/105046224944?origin=resultslist},
doi = {10.1109/CCECE68150.2026.11610144},
isbn = {08407789 (ISSN); 979-833158840-3 (ISBN)},
year = {2026},
date = {2026-01-01},
booktitle = {Can Conf Electr Comput Eng},
pages = {627–632},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Reliable crack segmentation is vital for structural health monitoring, yet the scarcity of well-annotated data constitutes a major challenge. To address this limitation, we propose a novel context-aware generative framework designed to synthesize realistic crack growth patterns for data augmentation. Unlike existing methods that primarily manipulate textures or background content, CrackForward explicitly models crack morphology by combining directional crack elongation with learned thickening and branching. Our framework integrates two key innovations: (i) a contextually guided crack expansion module, which uses local directional cues and adaptive random walk to simulate realistic propagation paths; and (ii) a two-stage U-Net-style generator that learns to reproduce spatially varying crack characteristics such as thickness, branching, and growth. Experimental results show that the generated samples preserve target-stage saturation and thickness characteristics and improve the performance of several crack segmentation architectures. These results indicate that structure-aware synthetic crack generation can provide more informative training data than conventional augmentation alone. © 2026 IEEE.},
note = {Journal Abbreviation: Can Conf Electr Comput Eng},
keywords = {Context-Aware, Crack Expansion, Crack morphology, Crack propagation, Crack segmentation, Crack segmentations, Data augmentation, Growth patterns, Health monitoring, Local Eigenvector Orientation, Morphology, Random processes, Structural health, Structural health monitoring, Textures, Unet-style generation},
pubstate = {published},
tppubtype = {inproceedings}
}
Reliable crack segmentation is vital for structural health monitoring, yet the scarcity of well-annotated data constitutes a major challenge. To address this limitation, we propose a novel context-aware generative framework designed to synthesize realistic crack growth patterns for data augmentation. Unlike existing methods that primarily manipulate textures or background content, CrackForward explicitly models crack morphology by combining directional crack elongation with learned thickening and branching. Our framework integrates two key innovations: (i) a contextually guided crack expansion module, which uses local directional cues and adaptive random walk to simulate realistic propagation paths; and (ii) a two-stage U-Net-style generator that learns to reproduce spatially varying crack characteristics such as thickness, branching, and growth. Experimental results show that the generated samples preserve target-stage saturation and thickness characteristics and improve the performance of several crack segmentation architectures. These results indicate that structure-aware synthetic crack generation can provide more informative training data than conventional augmentation alone. © 2026 IEEE.
2.
Zetout, A.; Allili, M. S.
CSDNet: Context-Aware Segmentation of Disaster Aerial Imagery Using Detection-Guided Features and Lightweight Transformers Journal Article
In: Remote Sensing, vol. 17, no. 14, 2025, ISSN: 20724292 (ISSN).
Abstract | Links | BibTeX | Tags: Aerial imagery, Affected area, Antennas, Class imbalance, Context-Aware, Contextual semantic segmentation, Contextual semantics, Detection, disaster response, Disaster-response, Emergency services, Error detection, Feature extraction, Lightweight model, Semantic segmentation, Semantics
@article{zetout_csdnet_2025,
title = {CSDNet: Context-Aware Segmentation of Disaster Aerial Imagery Using Detection-Guided Features and Lightweight Transformers},
author = {A. Zetout and M. S. Allili},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-105011677142&doi=10.3390%2Frs17142337&partnerID=40&md5=a83db334b208d065476e0026ad0ee416},
doi = {10.3390/rs17142337},
issn = {20724292 (ISSN)},
year = {2025},
date = {2025-01-01},
journal = {Remote Sensing},
volume = {17},
number = {14},
abstract = {Accurate multi-class semantic segmentation of disaster-affected areas is essential for rapid response and effective recovery planning. We present CSDNet, a context-aware segmentation model tailored to disaster scene scenarios, designed to improve segmentation of both large-scale disaster zones and small, underrepresented classes. The architecture combines a lightweight transformer module for global context modeling with depthwise separable convolutions (DWSCs) to enhance efficiency without compromising representational capacity. Additionally, we introduce a detection-guided feature fusion mechanism that integrates outputs from auxiliary detection tasks to mitigate class imbalance and improve discrimination of visually similar categories. Extensive experiments on several public datasets demonstrate that our model significantly improves segmentation of both man-made infrastructure and natural damage-related features, offering a robust and efficient solution for post-disaster analysis. © 2025 by the authors.},
keywords = {Aerial imagery, Affected area, Antennas, Class imbalance, Context-Aware, Contextual semantic segmentation, Contextual semantics, Detection, disaster response, Disaster-response, Emergency services, Error detection, Feature extraction, Lightweight model, Semantic segmentation, Semantics},
pubstate = {published},
tppubtype = {article}
}
Accurate multi-class semantic segmentation of disaster-affected areas is essential for rapid response and effective recovery planning. We present CSDNet, a context-aware segmentation model tailored to disaster scene scenarios, designed to improve segmentation of both large-scale disaster zones and small, underrepresented classes. The architecture combines a lightweight transformer module for global context modeling with depthwise separable convolutions (DWSCs) to enhance efficiency without compromising representational capacity. Additionally, we introduce a detection-guided feature fusion mechanism that integrates outputs from auxiliary detection tasks to mitigate class imbalance and improve discrimination of visually similar categories. Extensive experiments on several public datasets demonstrate that our model significantly improves segmentation of both man-made infrastructure and natural damage-related features, offering a robust and efficient solution for post-disaster analysis. © 2025 by the authors.
3.
Zetout, A.; Allili, M. S.
CSDNet: Context-Aware Segmentation of Disaster Aerial Imagery Using Detection-Guided Features and Lightweight Transformers Journal Article
In: Remote Sensing, vol. 17, no. 14, 2025, ISSN: 20724292 (ISSN).
Abstract | Links | BibTeX | Tags: Aerial imagery, Affected area, Antennas, Class imbalance, Context-Aware, Contextual semantic segmentation, Contextual semantics, Detection, disaster response, Disaster-response, Emergency services, Error detection, Feature extraction, Lightweight model, Semantic segmentation, Semantics
@article{zetoutCSDNetContextAwareSegmentation2025,
title = {CSDNet: Context-Aware Segmentation of Disaster Aerial Imagery Using Detection-Guided Features and Lightweight Transformers},
author = {A. Zetout and M. S. Allili},
url = {https://www.scopus.com/pages/publications/105011677142?origin=resultslist},
doi = {10.3390/rs17142337},
issn = {20724292 (ISSN)},
year = {2025},
date = {2025-01-01},
journal = {Remote Sensing},
volume = {17},
number = {14},
publisher = {Multidisciplinary Digital Publishing Institute (MDPI)},
abstract = {Accurate multi-class semantic segmentation of disaster-affected areas is essential for rapid response and effective recovery planning. We present CSDNet, a context-aware segmentation model tailored to disaster scene scenarios, designed to improve segmentation of both large-scale disaster zones and small, underrepresented classes. The architecture combines a lightweight transformer module for global context modeling with depthwise separable convolutions (DWSCs) to enhance efficiency without compromising representational capacity. Additionally, we introduce a detection-guided feature fusion mechanism that integrates outputs from auxiliary detection tasks to mitigate class imbalance and improve discrimination of visually similar categories. Extensive experiments on several public datasets demonstrate that our model significantly improves segmentation of both man-made infrastructure and natural damage-related features, offering a robust and efficient solution for post-disaster analysis. © 2025 by the authors.},
keywords = {Aerial imagery, Affected area, Antennas, Class imbalance, Context-Aware, Contextual semantic segmentation, Contextual semantics, Detection, disaster response, Disaster-response, Emergency services, Error detection, Feature extraction, Lightweight model, Semantic segmentation, Semantics},
pubstate = {published},
tppubtype = {article}
}
Accurate multi-class semantic segmentation of disaster-affected areas is essential for rapid response and effective recovery planning. We present CSDNet, a context-aware segmentation model tailored to disaster scene scenarios, designed to improve segmentation of both large-scale disaster zones and small, underrepresented classes. The architecture combines a lightweight transformer module for global context modeling with depthwise separable convolutions (DWSCs) to enhance efficiency without compromising representational capacity. Additionally, we introduce a detection-guided feature fusion mechanism that integrates outputs from auxiliary detection tasks to mitigate class imbalance and improve discrimination of visually similar categories. Extensive experiments on several public datasets demonstrate that our model significantly improves segmentation of both man-made infrastructure and natural damage-related features, offering a robust and efficient solution for post-disaster analysis. © 2025 by the authors.
4.
Zetout, A.; Allili, M. S.
CSDNet: Context-Aware Segmentation of Disaster Aerial Imagery Using Detection-Guided Features and Lightweight Transformers Journal Article
In: Remote Sensing, vol. 17, no. 14, 2025, ISSN: 20724292 (ISSN).
Abstract | Links | BibTeX | Tags: Aerial imagery, Affected area, Antennas, Class imbalance, Context-Aware, Contextual semantic segmentation, Contextual semantics, Detection, disaster response, Disaster-response, Emergency services, Error detection, Feature extraction, Lightweight model, Semantic segmentation, Semantics
@article{zetoutCSDNetContextAwareSegmentation2025a,
title = {CSDNet: Context-Aware Segmentation of Disaster Aerial Imagery Using Detection-Guided Features and Lightweight Transformers},
author = {A. Zetout and M. S. Allili},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-105011677142&doi=10.3390%2Frs17142337&partnerID=40&md5=a83db334b208d065476e0026ad0ee416},
doi = {10.3390/rs17142337},
issn = {20724292 (ISSN)},
year = {2025},
date = {2025-01-01},
journal = {Remote Sensing},
volume = {17},
number = {14},
abstract = {Accurate multi-class semantic segmentation of disaster-affected areas is essential for rapid response and effective recovery planning. We present CSDNet, a context-aware segmentation model tailored to disaster scene scenarios, designed to improve segmentation of both large-scale disaster zones and small, underrepresented classes. The architecture combines a lightweight transformer module for global context modeling with depthwise separable convolutions (DWSCs) to enhance efficiency without compromising representational capacity. Additionally, we introduce a detection-guided feature fusion mechanism that integrates outputs from auxiliary detection tasks to mitigate class imbalance and improve discrimination of visually similar categories. Extensive experiments on several public datasets demonstrate that our model significantly improves segmentation of both man-made infrastructure and natural damage-related features, offering a robust and efficient solution for post-disaster analysis. © 2025 by the authors.},
keywords = {Aerial imagery, Affected area, Antennas, Class imbalance, Context-Aware, Contextual semantic segmentation, Contextual semantics, Detection, disaster response, Disaster-response, Emergency services, Error detection, Feature extraction, Lightweight model, Semantic segmentation, Semantics},
pubstate = {published},
tppubtype = {article}
}
Accurate multi-class semantic segmentation of disaster-affected areas is essential for rapid response and effective recovery planning. We present CSDNet, a context-aware segmentation model tailored to disaster scene scenarios, designed to improve segmentation of both large-scale disaster zones and small, underrepresented classes. The architecture combines a lightweight transformer module for global context modeling with depthwise separable convolutions (DWSCs) to enhance efficiency without compromising representational capacity. Additionally, we introduce a detection-guided feature fusion mechanism that integrates outputs from auxiliary detection tasks to mitigate class imbalance and improve discrimination of visually similar categories. Extensive experiments on several public datasets demonstrate that our model significantly improves segmentation of both man-made infrastructure and natural damage-related features, offering a robust and efficient solution for post-disaster analysis. © 2025 by the authors.