Globally, the greatest concerns of farmers are plant diseases. A considerable loss of yield has an immediate impact on the economy. Machine learning models exhibited capabilities to detect plant diseases. To enhance the generalization, flexibility, and adaptability of machine learning models, an ensemble model was introduced to combine many models and produce an appropriate and optimal predictive solution. To achieve better accuracy in deep learning models, it is necessary to have a significant amount of training time and memory resources. This is frequently utilized in engineering fields. Nowadays, the computational burden of training multiple independent networks increases with the growing number of deep learning models. To address these computational challenges, we introduced pruning and tuning-based fast, cost-effective sparse ensemble networks. The popularity of sparse architecture has increased because of its low memory requirements and faster inference rates. To identify an appropriate subset of ensemble members to build a smaller ensemble model, pruning is used to refrain from retraining the model. To achieve this, we first trained a base network. After that, we clone the parent network and drastically reduce each child’s features to generate an ensemble of members with various topologies. We soon trained each successor network for a smaller number of epochs. An empirical investigation demonstrated how well our method works on the pruned ensemble model.