Gossypium herbaceum: Folium disease identification and classification using Efficient Net-Coordinate Convolutional Neural Network (EcoNet)

Globally, agriculture plays a key role in economic growth, with cotton being one of the most essential fiber crops, often referred to as “silver fiber” and widely used in manufacturing industries. However, cotton plants face productivity challenges due to disease outbreaks, making early detection crucial. Manual monitoring is tedious, and conventional cnn-based deep transfer learning models suffer from spatial information loss and translational variance, reducing classification accuracy. To address these challenges, this study proposes EcoNet, a novel Efficient Net-Coordinate Convolutional Neural Network (EcoNet) designed to enhance spatial feature extraction and positional encoding for improved cotton leaf disease identification and classification. The dataset was compiled using an amalgamated method: in-field fungal and healthy cotton leaf images were collected from Thoppulampatti, Madurai, India, while bacterial and insect-borne disease images were obtained from Kaggle's public repository. Pre-processing involved image resizing (224 × 224 pixels) and normalization to optimize model performance. EcoNet integrates Coordinate Convolution Neural Network (Coordconv) layers to capture local and global spatial dependencies, while residual blocks improve hierarchical feature extraction and reduce computational overhead. Experimental results demonstrate that EcoNet achieves 98.32 % accuracy with a Cohen's Kappa coefficient of 0.9804, outperforming conventional models.