HECNNet: Hybrid Ensemble Convolutional Neural Network Model with Multi-Backbone Feature Extractors for Soybean Disease Classification

The purpose of this study is to classify soybean plant diseases using a new method called Hybrid Ensemble Convolutional Neural Networks (HECNNet). To effectively extract unique characteristics and categorize pictures, the proposed approach makes use of convolutional neural networks. The suggested approach is organized around three main phases: data collection, preprocessing, and classification. The novel component of this study is the hybrid ensemble model, which consists of ResNeXt, SqueezeNet, and SegNet, three strong backbone networks. With convolutional layers, dropouts, and max-pooling layers carefully adjusted to match the needs of the dataset, these backbone networks function as feature extractors and classifiers. HECNNet is the product of the interplay between these three extractors. Notably, the prediction scores from each sub-network are combined using a weighted sum operation in HECNNet, which significantly improves the final recognition rate compared to a single network method. The proposed algorithm tackles the complexity of soybean disease categorization and obtains an amazing accuracy rate of 91.7%. This study is a major advance in agricultural disease diagnosis since it provides a reliable and effective method for diagnosing ailments affecting soybean plants.