SC-CNN: Advancing Corn Crop Disease Classification - A Comparative Study with CNN

In this study, the effectiveness and performance of the Skip-Connected Convolutional Neural Network (SC-CNN) architecture for image classification tasks are evaluated against that of a traditional Convolutional Neural Network (CNN). Inspired by the U-Net model, SC-CNN integrates skip connections to enable seamless feature propagation across layers, addressing challenges like vanishing gradients. Through a meticulous evaluation encompassing metrics, training efficiency, complexity analysis, and feature map visualization, we scrutinize SC-CNN’s effectiveness. Our findings highlight SC-CNN’s competitive performance and interpretability advantages. Notably, by achieving a 0.17% increase in accuracy, it appears that SC-CNN slightly outperforms CNN. This research fosters a deeper understanding of skip-connected architectures in image classification and their broader applicability in computer vision tasks.