← Back to Publications List

Comparative Analysis with Hybrid Model Design for Deep Learning Approaches in Detecting Plant Diseases

Students & Supervisors

Student Authors
Md Naved Akhter
Bachelor of Science in Computer Science & Engineering, FST
Riazul Zannah
Bachelor of Science in Computer Science & Engineering, FST
Sk. Nur Alam
Bachelor of Science in Computer Science & Engineering, FST
Supervisors
Sharfuddin Mahmood
Assistant Professor, Special Assistant [osa], FST

Abstract

Plant disease is a major world food security issue where early and reliable detection is paramount to preventing crop loss and sustainable farming. As much as there has been a lot of progress in Deep Learning, most papers are marred by unbalanced datasets or evaluation methods, creating a gap in fair performance comparison. To counteract this, we present a hybrid deep model where we join ResNet50, DenseNet201, and EfficientNet-B3 using transfer learning and incorporate fusion for robust plant disease classification on the PlantVillage dataset. The hybrid model proposed attained 98.91% accuracy with macro averaged precision, recall, and F1 score nearly reaching 99%, surpassing multiple individual CNN models. These results show the power of combining diverse feature extraction methodologies and proposing a robust benchmark for future studies. Our contribution demonstrates the expanded potential of hybrid architectures in building AI-based agricultural solutions.

Keywords

Hybrid CNN Architecture Transfer Learning Precision Agriculture PlantVillage dataset Smart Farm ing Feature Fusion Plant Disease

Publication Details

  • Type of Publication:
  • Conference Name: 28th International Conference on Computer and Information Technology (ICCIT 2025)
  • Date of Conference: 19/12/2025 - 19/12/2025
  • Venue: Long Beach Hotel Cox's Bazar Bangladesh
  • Organizer: IEEE
View on Publisher