DEEP TRANSFER LEARNING OPTIMIZATION USING HYBRID ENSEMBLE MODELS FOR MULTI-CLASS DISEASE DETECTION IN EGGPLANT LEAF VEGETABLE CROP
Shilpi Mishra1, Sandhya Sharma2
1Department of Computer Science and Engineering, 2Department of Electronics and Communication Engineering,
Suresh Gyan Vihar University, Jaipur, India
Abstract. Eggplant was one of the most widely grown vegetable crops in the Jaipur region and plays an important role in the income of local farmers. The eggplant vegetable crop was often affected by different leaf diseases such as insect pest infestation, leaf spot, mosaic virus, white mold, and wilt, which reduce yield and quality. Early and accurate recognition of these diseases was necessary for reducing losses and improving sustainable farming practices. In current study, different modern computer-based approaches were tested for identifying multiple eggplant diseases from leaf images. Several well-known image recognition models were first applied individually, but their accuracy remained low, ranging between 30% and 35%. A combined model using image features with an advanced decision system improved the accuracy to about 76%. Further, a voting-based approach that merges results from multiple models reached around 60% accuracy. The most effective solution was a stacking-based hybrid system that combined the strengths of all models. Present approach achieved a very high accuracy of 99.8%, with precision, recall, F1-score, and area under curve values also close to 1.0. The statistical analysis confirmed that current hybrid method was significantly better than other tested methods, with Wilcoxon signed-rank tests showing clear improvements in most pairwise comparisons. Sample predictions further demonstrated that the hybrid model was able to correctly identify almost all cases of healthy and diseased leaves, while other models often confused between classes. These results show that a carefully designed hybrid system can provide reliable disease recognition in eggplant crops. The proposed method can help farmers and extension workers in Jaipur to monitor plant health more effectively and reduce the economic losses caused by leaf diseases, supporting sustainable vegetable production in the region.
Keywords: Eggplant leaf diseases, Hybrid detection framework, Deep transfer learning, Ensemble classification, Sustainable agriculture.
VOLUME 10 ISSUE 02 2026: 13 – 23