| G. Vonitsanos, E.-E. Economopoulou, S. Sioutas, A. Kanavos, Ph. Mylonas |
| Enhancing Image Classification with Attention-Driven Convolutional Neural Networks on CIFAR Datasets |
| 10th South-East Europe Design Automation, Computer Engineering, Computer Networks and Social Media Conference (SEEDA-CECNSM 2025), Patras, Greece, September 19-21, 2025 |
|
ABSTRACT
|
| Image classification is a core task in computer vision with wide-ranging applications, from autonomous vehicles to medical diagnostics. While Convolutional Neural Networks (CNNs) have demonstrated strong performance by learning spatial hierarchies of features, they often struggle to capture complex interdependencies among spatial and channel-wise representations. This work proposes an attention-augmented CNN architecture that integrates both Squeeze-and-Excitation (SE) and Convolutional Block Attention Module (CBAM) mechanisms to enhance feature selection and improve classification performance. The proposed model is evaluated on two benchmark datasets, CIFAR-10 and CIFAR-100, and compared against baseline CNN and Artificial Neural Network (ANN) architectures. Experimental results indicate that the attention-enhanced CNN achieves superior classification accuracy and generalization, with notable gains in distinguishing visually similar classes. These findings highlight the effectiveness of combining channel and spatial attention modules to improve the robustness and adaptability of deep learning-based visual recognition systems.
|
| 19 September, 2025 |
| G. Vonitsanos, E.-E. Economopoulou, S. Sioutas, A. Kanavos, Ph. Mylonas, "Enhancing Image Classification with Attention-Driven Convolutional Neural Networks on CIFAR Datasets", 10th South-East Europe Design Automation, Computer Engineering, Computer Networks and Social Media Conference (SEEDA-CECNSM 2025), Patras, Greece, September 19-21, 2025 |
[  PDF] [
BibTex] [
Print] [
 Back] |
