Plant diseases pose a major threat to global food security by reducing crop yield and quality. Early and accurate diagnosis is essential for timely intervention, yet traditional manual inspection methods are slow, subjective, and impractical for large-scale monitoring. Recent advances in deep learning, particularly Convolutional Neural Networks (CNNs), have significantly improved plant disease detection, but conventional CNN architectures face challenges such as high computational cost, overfitting with limited data, and difficulty in capturing multi-scale disease features. This study proposes an Enhanced Convolutional Neural Network (ECNN) for accurate leaf-based plant disease classification. The proposed architecture integrates multi-scale feature extraction blocks (MSFEB), channel and spatial attention mechanisms (CBAM), and depthwise separable convolutions to enhance feature representation while maintaining efficiency. Experiments were conducted on the PlantVillage dataset supplemented with field-collected and augmented images, covering multiple crop species and disease categories. The ECNN achieved superior performance compared to baseline models, with 98.7% accuracy, 98.8% precision, 98.7% recall, and 98.65% F1-score, outperforming VGG16, ResNet50, MobileNetV2, and EfficientNet-B0. In addition, ECNN maintained a lightweight architecture with only 4.9M parameters and an inference time of 19 ms per image, demonstrating its suitability for real-time deployment on edge devices in agricultural environments.