Anouar Chahbouni
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Anouar Chahbouni has a total of 14 co-authors. Here are the top co-authors by number of shared publications:
- Badre Bossoufi
LIMAS Laboratory, Faculty of Sciences Dhar El Mahraz, Sidi Mohamed Ben Abdellah University, Fez 30050, Morocco
3 shared publications - Imane El Manaa
Faculty of Sciences Dhar El Mehraz, Sidi Mohammed Ben Abdellah University, Fez 30000, Morocco
3 shared publications - Khaoula El Manaa
Faculty of Sciences Dhar El Mehraz, Sidi Mohammed Ben Abdellah University, Fez 30000, Morocco
3 shared publications - Mohammed El Ghzaoui
Faculty of Sciences Dhar El Mehraz, Sidi Mohammed Ben Abdellah University, Fez 30000, Morocco
3 shared publications - Rachid El Alami
Faculty of Sciences Dhar El Mehraz, Sidi Mohammed Ben Abdellah University, Fez 30000, Morocco
3 shared publications - Yassine Abouch
DAKAI Laboratory, Nextronic by Aba Technology, Casablanca 20253, Morocco
3 shared publications
Latest Publications
- Open Access
Beyond Feature Similarity: A Local and Global Redundancy Analysis
Deep neural networks achieve remarkable performance in visual recognition tasks, but this accuracy often comes at the cost of high computational complexity and over-parameterized representations. Structured pruning has emerged as an effective solution for reducing this complexity by eliminating unnecessary representations. Nevertheless, its success implicitly relies on the existence of exploitable redundancy within learned feature spaces, whereas the nature, structure, and evolution of such redundancy remain largely underexplored. To address this gap, this paper presents a comprehensive redundancy analysis framework that investigates deep representations from complementary visual, geometric, and intrinsic dimensionality perspectives. The framework is evaluated using VGG16 and Swin-Tiny on the CIFAR-10 and PlantVillage datasets, enabling the analysis of redundancy across convolutional and transformer-based architectures, multiple representation levels, and distinct image classification domains. The proposed framework combines feature-map visualization, cosine similarity analysis, singular value decomposition (SVD), and effective rank (ER) to jointly characterize feature redundancy, representational dependencies, and intrinsic dimensionality from complementary local and global perspectives. The results reveal that redundancy is a pervasive yet dynamic property of deep neural networks, evolving according to architecture, dataset characteristics, representation depth, and input distributions. Pairwise redundancy ratios reach up to 18.61% in deep representations, whereas normalized effective rank values decrease to as low as 0.01, indicating strong spectral concentration in some representations. More importantly, the study demonstrates that redundancy extends beyond directly similar features and also encompasses global dependencies and representations occupying reduced effective subspaces. Structured pruning validation further demonstrates substantial compression with preserved predictive performance under the evaluated experimental settings. These findings provide a deeper understanding of learned representations, establish a principled perspective for interpreting pruning effectiveness, and offer practical guidelines for developing future redundancy-aware compression methods for efficient and deployable deep learning systems.
Open AccessBeyond Feature Similarity: A Local and Global Redundancy Analysis
Deep neural networks achieve remarkable performance in visual recognition tasks, but this accuracy often comes at the cost of high computational complexity and over-parameterized representations. Structured pruning has emerged as an effective solution for reducing this complexity by eliminating unnecessary representations. Nevertheless, its success implicitly relies on the existence of exploitable redundancy within learned feature spaces, whereas the nature, structure, and evolution of such redundancy remain largely underexplored. To address this gap, this paper presents a comprehensive redundancy analysis framework that investigates deep representations from complementary visual, geometric, and intrinsic dimensionality perspectives. The framework is evaluated using VGG16 and Swin-Tiny on the CIFAR-10 and PlantVillage datasets, enabling the analysis of redundancy across convolutional and transformer-based architectures, multiple representation levels, and distinct image classification domains. The proposed framework combines feature-map visualization, cosine similarity analysis, singular value decomposition (SVD), and effective rank (ER) to jointly characterize feature redundancy, representational dependencies, and intrinsic dimensionality from complementary local and global perspectives. The results reveal that redundancy is a pervasive yet dynamic property of deep neural networks, evolving according to architecture, dataset characteristics, representation depth, and input distributions. Pairwise redundancy ratios reach up to 18.61% in deep representations, whereas normalized effective rank values decrease to as low as 0.01, indicating strong spectral concentration in some representations. More importantly, the study demonstrates that redundancy extends beyond directly similar features and also encompasses global dependencies and representations occupying reduced effective subspaces. Structured pruning validation further demonstrates substantial compression with preserved predictive performance under the evaluated experimental settings. These findings provide a deeper understanding of learned representations, establish a principled perspective for interpreting pruning effectiveness, and offer practical guidelines for developing future redundancy-aware compression methods for efficient and deployable deep learning systems.
- Open Access
Adaptation-Preserving Structured Pruning for Efficient Transfer Learning in Deep Neural Networks
Transfer learning enables pretrained deep networks to adapt to new tasks but typically preserves their original computational structure, leaving much of the inference cost unchanged. Structured pruning can reduce this complexity; however, introducing sparsity during fine-tuning creates a competing optimization problem in which transferred representations must adapt while the network simultaneously determines which structures can be removed. We propose GASP, an adaptation-preserving structured pruning framework that treats task adaptation and structural sparsity learning as a coordinated optimization process, allowing for the sparse structure to emerge progressively as pretrained representations adapt to the target task. Rather than imposing structural decisions independently of this adaptation, GASP first establishes target-task adaptation through a warm-up phase, before progressively introducing sparsity pressure through an equation-driven schedule. In parallel, temperature annealing gradually sharpens input-aware continuous gates toward increasingly differentiated structural responses. Together, these mechanisms regulate the onset and progression of sparsification while structural importance evolves during fine-tuning. The learned gating responses are subsequently aggregated over the validation data and converted into a physically compact architecture through global Otsu-based thresholding and structural reconstruction, avoiding predefined sparsity ratios and iterative prune–retrain cycles. The same optimization principle is applied to channels in CNNs and MLP units in Swin Transformers without additional trainable mask parameters. Experiments with VGG16, EfficientNet-B0, and Swin-Tiny on CIFAR-10, PlantVillage, and PlantDoc show that GASP achieves accuracy improvements of up to 2.22%, with the corresponding configuration reducing parameters and FLOPs by 91.37% and 88.83%, respectively. Raspberry Pi 4 deployment further demonstrates that the learned structural sparsity translates into practical inference efficiency. Overall, the results demonstrate the effectiveness of coordinating representation adaptation and structural sparsification when deriving compact task-adapted models.
Open AccessAdaptation-Preserving Structured Pruning for Efficient Transfer Learning in Deep Neural Networks
Transfer learning enables pretrained deep networks to adapt to new tasks but typically preserves their original computational structure, leaving much of the inference cost unchanged. Structured pruning can reduce this complexity; however, introducing sparsity during fine-tuning creates a competing optimization problem in which transferred representations must adapt while the network simultaneously determines which structures can be removed. We propose GASP, an adaptation-preserving structured pruning framework that treats task adaptation and structural sparsity learning as a coordinated optimization process, allowing for the sparse structure to emerge progressively as pretrained representations adapt to the target task. Rather than imposing structural decisions independently of this adaptation, GASP first establishes target-task adaptation through a warm-up phase, before progressively introducing sparsity pressure through an equation-driven schedule. In parallel, temperature annealing gradually sharpens input-aware continuous gates toward increasingly differentiated structural responses. Together, these mechanisms regulate the onset and progression of sparsification while structural importance evolves during fine-tuning. The learned gating responses are subsequently aggregated over the validation data and converted into a physically compact architecture through global Otsu-based thresholding and structural reconstruction, avoiding predefined sparsity ratios and iterative prune–retrain cycles. The same optimization principle is applied to channels in CNNs and MLP units in Swin Transformers without additional trainable mask parameters. Experiments with VGG16, EfficientNet-B0, and Swin-Tiny on CIFAR-10, PlantVillage, and PlantDoc show that GASP achieves accuracy improvements of up to 2.22%, with the corresponding configuration reducing parameters and FLOPs by 91.37% and 88.83%, respectively. Raspberry Pi 4 deployment further demonstrates that the learned structural sparsity translates into practical inference efficiency. Overall, the results demonstrate the effectiveness of coordinating representation adaptation and structural sparsification when deriving compact task-adapted models.
- 1 May 2026
- Published by Springer Nature
- p. 341-350
A Lightweight CNN-Based Model for Early Tomato Diseases Classification
- 1 May 2026
- Published by Springer Nature
- p. 341-350
A Lightweight CNN-Based Model for Early Tomato Diseases Classification
- Open Access
- 24 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 5358-5373
RyNet: Multi-Level Attention for Rich and Complementary Feature Representation in CNNs
Convolutional neural networks (CNNs) excel at extracting spatial and channel-wise features, but reliance on final-layer activations often underutilizes intermediate representations and weakens hierarchical cues. We propose RyNet, a theoretically grounded, modular, and CNN-agnostic dual-attention framework that integrates dynamic channel refinement with multi-level feature fusion in a single backbone. RyNet aggregates low-, mid-, and high-level feature maps, applies channel squeezing with learnable channel-wise attention scores to promote selective activation, and fuses recalibrated features via a lightweight summation strategy. This yields stage-to-stage coupling with implicit spatial emphasis at negligible computational cost. RyNet achieves hierarchical complementarity without branch duplication or concatenation–projection stacks. Integrated into VGG16 and ResNet18, RyNet improves CIFAR-10 test accuracy from 89.90% to 90.84% (a relative improvement of 1.04%) and from 90.42% to 91.74% (a relative improvement of 1.50%), respectively, On the PlantVillage (tomato) dataset, a custom CNN-based model integrated with RyNet achieves 99.61% accuracy, representing a relative improvement of 1.54%. Under identical datasets and training conditions, RyNet consistently and statistically outperforms SE and CBAM, and beyond this, demonstrates superiority over recent CNN–Transformer hybrids and other state-of-the-art methods. Feature-space analysis reveals a 121.9% increase in inter-class separability, while visualizations and interpretability evaluations confirm the effectiveness of dynamic channel selection under multi-level fusion. These results provide strong theoretical, empirical, and statistical validation for the proposed design. Overall, RyNet emerges as a principled, lightweight, and easily integrable module for enhancing CNN representations in resource-constrained environments, with significant performance gains and improved model explainability.
- 24 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 5358-5373
Open AccessRyNet: Multi-Level Attention for Rich and Complementary Feature Representation in CNNs
Convolutional neural networks (CNNs) excel at extracting spatial and channel-wise features, but reliance on final-layer activations often underutilizes intermediate representations and weakens hierarchical cues. We propose RyNet, a theoretically grounded, modular, and CNN-agnostic dual-attention framework that integrates dynamic channel refinement with multi-level feature fusion in a single backbone. RyNet aggregates low-, mid-, and high-level feature maps, applies channel squeezing with learnable channel-wise attention scores to promote selective activation, and fuses recalibrated features via a lightweight summation strategy. This yields stage-to-stage coupling with implicit spatial emphasis at negligible computational cost. RyNet achieves hierarchical complementarity without branch duplication or concatenation–projection stacks. Integrated into VGG16 and ResNet18, RyNet improves CIFAR-10 test accuracy from 89.90% to 90.84% (a relative improvement of 1.04%) and from 90.42% to 91.74% (a relative improvement of 1.50%), respectively, On the PlantVillage (tomato) dataset, a custom CNN-based model integrated with RyNet achieves 99.61% accuracy, representing a relative improvement of 1.54%. Under identical datasets and training conditions, RyNet consistently and statistically outperforms SE and CBAM, and beyond this, demonstrates superiority over recent CNN–Transformer hybrids and other state-of-the-art methods. Feature-space analysis reveals a 121.9% increase in inter-class separability, while visualizations and interpretability evaluations confirm the effectiveness of dynamic channel selection under multi-level fusion. These results provide strong theoretical, empirical, and statistical validation for the proposed design. Overall, RyNet emerges as a principled, lightweight, and easily integrable module for enhancing CNN representations in resource-constrained environments, with significant performance gains and improved model explainability.
- Open Access
- 29 September 2025
- Published by MDPI AG in Machine Learning and Knowledge Extraction
- Vol. 7 (4) , 110
Attention-Guided Differentiable Channel Pruning for Efficient Deep Networks
Deploying deep learning (DL) models in real-world environments remains a major challenge, particularly under resource-constrained conditions where achieving both high accuracy and compact architectures is essential. While effective, Conventional pruning methods often suffer from high computational overhead, accuracy degradation, or disruption of the end-to-end training process, limiting their practicality for embedded and real-time applications. We present Dynamic Attention-Guided Pruning (DAGP), a Dynamic Attention-Guided Soft Channel Pruning framework that overcomes these limitations by embedding learnable, differentiable pruning masks directly within convolutional neural networks (CNNs). These masks act as implicit attention mechanisms, adaptively suppressing non-informative channels during training. A progressively scheduled L1 regularization, activated after a warm-up phase, enables gradual sparsity while preserving early learning capacity. Unlike prior methods, DAGP is retraining-free, introduces minimal architectural overhead, and supports optional hard pruning for deployment efficiency. Joint optimization of classification and sparsity objectives ensures stable convergence and task-adaptive channel selection. Experiments on CIFAR-10 (VGG16, ResNet56) and PlantVillage (custom CNN) achieve up to 98.82% FLOPs reduction with accuracy gains over baselines. Real-world validation on an enhanced PlantDoc dataset for agricultural monitoring achieves 60 ms inference with only 2.00 MB RAM on a Raspberry Pi 4, confirming efficiency under field conditions. These results illustrate DAGP’s potential to scale beyond agriculture to diverse edge-intelligent systems requiring lightweight, accurate, and deployable models.
- 29 September 2025
- Published by MDPI AG in Machine Learning and Knowledge Extraction
- Vol. 7 (4) , 110
Open AccessAttention-Guided Differentiable Channel Pruning for Efficient Deep Networks
Deploying deep learning (DL) models in real-world environments remains a major challenge, particularly under resource-constrained conditions where achieving both high accuracy and compact architectures is essential. While effective, Conventional pruning methods often suffer from high computational overhead, accuracy degradation, or disruption of the end-to-end training process, limiting their practicality for embedded and real-time applications. We present Dynamic Attention-Guided Pruning (DAGP), a Dynamic Attention-Guided Soft Channel Pruning framework that overcomes these limitations by embedding learnable, differentiable pruning masks directly within convolutional neural networks (CNNs). These masks act as implicit attention mechanisms, adaptively suppressing non-informative channels during training. A progressively scheduled L1 regularization, activated after a warm-up phase, enables gradual sparsity while preserving early learning capacity. Unlike prior methods, DAGP is retraining-free, introduces minimal architectural overhead, and supports optional hard pruning for deployment efficiency. Joint optimization of classification and sparsity objectives ensures stable convergence and task-adaptive channel selection. Experiments on CIFAR-10 (VGG16, ResNet56) and PlantVillage (custom CNN) achieve up to 98.82% FLOPs reduction with accuracy gains over baselines. Real-world validation on an enhanced PlantDoc dataset for agricultural monitoring achieves 60 ms inference with only 2.00 MB RAM on a Raspberry Pi 4, confirming efficiency under field conditions. These results illustrate DAGP’s potential to scale beyond agriculture to diverse edge-intelligent systems requiring lightweight, accurate, and deployable models.