Attention-Enhanced Deep Convolutional Neural Networks for Multi-Scale Feature Representation in Complex Image Classification

Abstract

Complex image classification is a key research area in computer vision with applications in medical imaging, autonomous vehicles, remote sensing, surveillance, industrial automation, and multimedia analysis. Convolutional Neural Networks (CNNs) have achieved strong performance by learning hierarchical feature representations from images. However, traditional CNNs struggle to capture multi-scale spatial information and long-range contextual dependencies, especially in complex images with variations in scale, illumination, occlusion, texture, and background clutter. These limitations reduce classification accuracy and affect robustness in real-world dynamic environments. To address these challenges, this research proposes an Attention-Enhanced Deep Convolutional Neural Network (AEDCNN) for multi-scale feature representation in complex image classification. The proposed model integrates deep convolutional layers with spatial and channel attention mechanisms to improve feature extraction and localization. Multi-scale convolution operations are used to capture both fine-grained local patterns and high-level semantic information across different resolutions. Attention modules help the network focus on important regions while suppressing irrelevant background noise, enhancing discriminative learning. The architecture further incorporates residual connections and feature fusion strategies to improve representation efficiency and stability. The AEDCNN model is evaluated on benchmark datasets using accuracy, precision, recall, F1-score, computational efficiency, and robustness metrics. Experimental results show that the proposed framework outperforms traditional CNN and existing deep learning models. Overall, the proposed approach enables more accurate, scalable, and robust image classification for complex visual environments and advanced AI-based vision systems.