Comparative Analysis of Deep Learning Models  for Tomato Leaf Disease Classification Using  Transfer Learning

Akshara R; Thilagaraj T; Arathi Sudarshan

doi:10.36948/ijfmr.2026.v08i03.80069

Comparative Analysis of Deep Learning Models for Tomato Leaf Disease Classification Using Transfer Learning

Author(s)	Ms. Akshara R, Dr. Thilagaraj T, Dr. Arathi Sudarshan
Country	India
Abstract	Tomato cultivation is an important component of global agricultural output, yet the crop remains highly susceptible to leaf diseases such as Bacterial Spot and Early Blight, both of which cause measurable losses in yield and produce quality when left undetected. Conventional approaches to disease identification rely on manual observation by trained personnel, a method that is inherently slow, subjective, and difficult to scale across large farming operations. This study investigates whether deep learning, applied through a transfer learning framework, can provide a reliable automated alternative.Three convolutional neural network architectures - MobileNetV2, EfficientNetB0, and VGG16 were evaluated and compared under identical experimental conditions. A dataset of 4,718 tomato leaf images, distributed across three classes — Bacterial Spot (2,127 images), Early Blight (1,000 images), and Healthy (1,591 images) was used for training and validation. Each image was resized to 224 x 224 pixels, normalized to the [0, 1] range, and augmented using horizontal flipping, zoom, and shear transformations. Pre-trained ImageNet weights were loaded for each base model, with the original classification layers replaced by a custom head consisting of Global Average Pooling, Dense (256 units, ReLU), Dropout (0.5), and Softmax output layers. All models were compiled with the Adam optimizer at a learning rate of 0.0001 and trained with categorical cross-entropy loss over an 80:20 train-validation split.EfficientNetB0 achieved the highest validation accuracy of 98.83% with the lowest validation loss of 0.0301. MobileNetV2 reached 97.77% validation accuracy and VGG16 achieved 98.41%, with all three models delivering an overall classification accuracy of approximately 98%. Early Blight was found to be the most challenging class across all models, owing to its visually subtle and variable symptom presentation. The findings suggest that EfficientNetB0 offers the best balance between accuracy and generalization, while MobileNetV2 is better suited to real-time deployment in resource-constrained environments. This study provides a reproducible, controlled benchmark for architecture selection in tomato disease detection applications
Keywords	Tomato Leaf Disease, Deep Learning, Transfer Learning, MobileNetV2, EfficientNetB0, VGG16, Convolutional Neural Network, Image Classification, Precision Agriculture
Field	Computer > Artificial Intelligence / Simulation / Virtual Reality
Published In	Volume 8, Issue 3, May-June 2026
Published On	2026-06-06
DOI	https://doi.org/10.36948/ijfmr.2026.v08i03.80069

View / Download PDF File

E-ISSN 2582-2160

doi

CrossRef DOI prefix of IJFMR is 10.36948/ijfmr

Downloads

Research Paper Format Copyright Permission Form and Undertaking Form Cover Page Vol 8 Isu 4 Cover Page Vol 8 Isu 3 Cover Page Vol 8 Isu 2

All research papers published on this website are licensed under Creative Commons Attribution-ShareAlike 4.0 International License, and all rights belong to their respective authors/researchers.

CC-BY-SA

About IJFMR Fees & Payment Current Issue Publication Archive	Submit Research Paper Track Submission Status Publication Guidelines Publication Ethics Peer Review & Plagiarism	Join as a Reviewer Editors & Reviewers Reviewer Referral Program Get Reviewer Membership Certi.	Website/Journal Policies Usage Policy Content Policies Privacy Policy

Contact Us		+91-9687-828-838	editor@ijfmr.com

International Journal For Multidisciplinary Research

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Bi-monthly Scholarly International Journal

Comparative Analysis of Deep Learning Models for Tomato Leaf Disease Classification Using Transfer Learning

Share this