Introduction:

Tomato is a fascinating plant that has captured the attention of botanists, agriculturalists, and culinary
enthusiasts alike. As a short-duration crop that yields highly, tomato production is an important
economic activity, with its cultivation increasing worldwide. tomato is a widely consumed vegetable,
eaten raw, as an ingredient in many dishes, sauces, salads, and even drinks. Advances in modern
agricultural technologies have enabled the production of sufficient food to meet global demands,
significantly impacting India's economy. Despite the popularity and economic importance of tomato,
maintaining the quality and yield of this crop has presented some challenges. Tomato production
requires a high level of management, labour, and capital, and is susceptible to variations in weather
that can result in severe crop damage and losses. However, food safety remains a major concern due
to factors such as climate change, decreasing pollinator numbers, plant diseases, and other challenges.
The nutritional value of tomato, particularly its vitamin content, has made it a popular item on menus
globally. Tomatoes, which are widely cultivated and highly nutritious, are a vital component of global
agricultural production and trade. According to the Food and Agriculture Organization of the United
Nations, tomato production has experienced substantial growth over the years, highlighting its
importance among vegetable crops. Nonetheless, diseases continue to pose a significant threat to
tomato production, leading to considerable economic losses in the agricultural sector. Diseases such as
early blight and late blight can cause severe reductions in crop yields. In India, smallholder farmers,
who contribute over 80 percent of agricultural output, often suffer from up to 50 percent of crop yield
losses due to diseases. Historically, plant diseases have been diagnosed through visual inspection, a
method that lacks precision and is limited by the scarcity of experts in rural areas. Early detection of
plant diseases is crucial for implementing preventive measures against pathogens. The ability to
rapidly and accurately identify tomato leaf diseases through the use of computer vision and machine
learning techniques holds great promise for the future of sustainable tomato production. By enabling
early detection and targeted treatment of diseases, these technologies can help to minimize crop losses
and ensure a reliable supply of this important vegetable to consumers worldwide. By using automated
systems to classify plant diseases through leaf images, a more accurate and efficient method than
traditional visual inspections can be achieved. This technological approach enhances diagnostic
accuracy and allows for timely intervention, improving disease management and crop protection
strategies. [1] Automatic detection of tomato leaf diseases is crucial for timely intervention and
minimizing agricultural losses.. This paper proposed a novel approach for identifying tomato leaf
diseases using a lightweight shared-twin neural network architecture. This method addressed
traditional algorithms' limitations by offering a computationally efficient solution suitable for
deployment in resource-constrained environments. The effectiveness of the proposed method will be
evaluated on benchmark datasets, comparing its performance to existing state-of-the-art approaches.
[2] (Sunil et al., 2023). This article presented a novel approach for tomato plant disease classification
using a Multilevel Feature Fusion Network. The proposed MFFN leverages the strengths of ResNet50
for feature extraction, combined with an adaptive attention mechanism that focuses on important
channel, spatial, and pixel-level information. The authors demonstrated the effectiveness of their
approach on a tomato plant leaf image dataset, achieving state-of-the-art classification accuracy. [3]
(Kumar et al., 2024) highlighted the significance of early and precise disease detection in tomato
production for global food security. It emphasized the use of deep learning, particularly YOLO
models, combined with transfer and ensemble learning techniques, to achieve highly accurate
identification and localization of tomato leaf diseases. The authors aimed to offer a practical solution
for real-time disease diagnosis, enabling timely intervention by farmers and ultimately contributing to
improved tomato yield. [4] This research involved eleven classes of tomato leaf diseases, including
one healthy class, and utilized data augmentation techniques to enhance the performance of the
models. The proposed custom CNN model achieved the highest accuracy and recall of 95.00% among
all the models tested [5] The proposed model integrates residual blocks and attention extraction
modules to accurately identify complex features of various tomato leaf diseases. The study aimed to
enhance the accuracy and efficiency of disease diagnosis in tomato plants, which is crucial for yield
and agricultural productivity.[6] contributions include the design of a tomato leaf disease
identification model based on YOLOX-MobileNetV3, the introduction of a sample adaptive cross-
entropy loss function, and the successful deployment of the enhanced model on an embedded device,
showcasing its practicality and feasibility for real-world applications in agricultural disease
identification.[7] The proposed pipeline includes image preparation, CNN training and feature
extraction with transfer learning, feature concatenation and selection, and classification. The
evaluation metrics used are precision, specificity, F1-score, accuracy, sensitivity, and Matthew
correlation coefficient (MCC). [8] utilized a dataset of tomato leaf images and employed various
machine learning algorithms, including Random Forest, KNN, and Naïve Bayes, to extract features
and classify diseases. The Random Forest algorithm achieved the highest accuracy of 98% among the
8 algorithms tested, correctly identifying 10 different diseases.. [9] The proposed model utilizes a
modified Xception architecture, which is pre-trained on the ImageNet dataset and then fine-tuned with
a dataset of tomato leaf images.[10] focuses on the application of three pre-trained CNNs—VGG-19,
Xception, and ResNet-50—to identify tomato plant diseases from leaf images. The study aims to
enhance agricultural practices by providing early diagnosis and treatment of plant diseases, thereby
reducing yield losses.[11] (Chaitanya et al., 2022) found that their Convolutional Neural Network
model achieved a mean accuracy of 91.2% in identifying nine common tomato leaf diseases and
healthy leaves. Essentially, they created a model that can accurately detect diseases in tomato plants
just by "looking" at images of their leaves. [12] (Mezenner et al., 2022) proposed a system that
combines Convolutional Neural Networks for feature extraction and Support Vector Machines for
classification. [13] The developed system achieved an accuracy of 95% in classifying tomato leaves
as healthy or diseased. The CNN model outperformed the SVM model in terms of accuracy and
computational efficiency including bacterial spot, early blight, and powdery mildew.[14] The
Inception structure allows for multi-scale feature fusion, enhancing accuracy on the leaf disease
dataset. The global average pooling layer replaces the fully connected layer, reducing the number of
model parameters and training time.[15] The proposed SEDCNN architecture integrates standard
Squeeze and Excitation (SE) blocks and achieved a high disease identification accuracy of 97%,
which is further improved to 98.86% with data augmentation.The research addresses the challenge of
chili leaf disease identification due to its significant impact on the yield and quality of the crop..[ 16]
The proposed method, termed adversarial class-incremental learning (ACiL), employs a rehearsal-
based approach to store and replay exemplars of previously learned classes. An improved adversarial
training algorithm is used to finetune the model on new classes without forgetting the old ones. The
method is evaluated on three datasets: PlantLeaf1, PlantLeaf2, and PlantLeaf3, which include IID and
OOD images. [17] developed an automated system for the detection and classification of potato leaf
diseases using digital image processing and machine learning techniques. The high accuracy rates for
early blight and healthy leaves, along with the satisfactory performance for late blight, indicate the
system's potential for early disease detection and management, contributing to increased potato
productivity and profitability for farmers.[18] (Vishnoi et al., 2023) The authors utilized data
augmentation techniques like shift, shear, scaling, zoom, and flipping to increase the training data
size. They trained and evaluated their model using the PlantVillage dataset, focusing on identifying
apple scab, black rot, and cedar rust diseases. [19] (Wang et al., 2024) The study highlighted the
effectiveness of this model in achieving high-precision identification of apple leaf diseases, even with
complex backgrounds.[20] The HOWSVD-MDA method proposed in (OUAMANE et al., 2016) is a
two-phase approach for tomato leaf disease classification. Phase 1: HOWSVD-This phase focused on
feature extraction. Phase 2: MDA-With the most relevant features extracted, MDA comes into play.
[21] authors enhanced multi-scale feature fusion by adding a fourth prediction layer, they introduced
the APEIOU loss function to address overlapping bounding boxes and improve accuracy. [22] The
method addressed the challenges of segmenting diseased leaf parts in varying rice growing
environments and complex disease features. The AISOA-SSformer introduced three key innovations:
the Sparse Global-Update Perceptron (SGUP) for real-time parameter updating, the Salient Feature
Attention Mechanism (SFAM) with spatial and channel reconstruction modules (SRM and CRM) for
feature extraction, and the Annealing-Integrated Sparrow Optimization Algorithm (AISOA) for fine-
tuning the model. The SGUP enhances model stability by dynamically adjusting weights, while the
SFAM focuses on salient features and reduced background interference. The AISOA improved
adaptability and robustness, particularly for fuzzy edge features. The method outperforms other
algorithms on a homemade dataset, achieving an 83.1% Mean Intersection Over Union (MIoU), an
80.3% Dice coefficient, and a 76.5% recall, with a model size of only 14.71 M.[23] The research
successfully demonstrated the potential of using classification tree models for the automated detection
and classification of tomato plant diseases. By leveraging digital image processing and machine
learning techniques, the study provided a valuable tool for enhancing agricultural disease management
practices. Future work may explore the application of more sophisticated techniques, such as neural
networks and support vector machines, for further improving classification accuracy. [24] The study
employed a CNN model for disease detection in tomato leaves. The dataset, sourced from Plant
Village, includes over 50,000 images of 14 crops, with a focus on tomato leaves exhibiting nine
different diseases. Data augmentation techniques were used to balance the dataset, and the model was
trained on a NVIDIA DGX v100 machine. The architecture features three convolution and max
pooling layers, with varying numbers of filters, followed by two fully connected layers. The proposed
CNN model achieved a testing accuracy ranging from 76% to 100% across the different disease
classes, with an average accuracy of 91.2%. This performance surpassed that of pre-trained models
like VGG16, InceptionV3, and Mobile Net, which had lower accuracies and higher storage
requirements. The model's efficiency and accuracy make it suitable for mobile devices with limited
storage, offering a practical solution for farmers. [25] The study introduced a method for local and
global class annotation and data augmentation to enhance accuracy and reduce false positives. The
system is trained and tested on a large Tomato Diseases and Pests Dataset, which includes images
with diverse conditions and variations. The experimental results demonstrated the system's
effectiveness in recognizing nine different types of diseases and pests, even in complex scenarios. The
authors highlighted the advantages of their approach over traditional methods, which include in-place
image capture, avoiding laboratory analysis, and the ability to handle multiple diseases or pests in one
sample. The system is designed to work with images taken by various camera devices, making it
practical for real-time field use without expensive technology. [26] indicated that the VGG model
outperformed others with a top-1 error rate of only 0.47%, also explored the impact of training image
type (laboratory vs. field conditions) on model performance, finding that models trained on field
condition images are more effective in identifying laboratory condition images than vice versa. by
emphasizing the potential of deep learning models for plant disease identification and the need for a
diverse and extensive database to improve the robustness and applicability of such systems. [ 27]
propose a novel convolutional neural network (CNN) model named EfficientNet-MG, which
enhanced the conventional EfficientNet by incorporating multistage feature fusion (MSFF) and the
Gaussian error linear unit (GELU) activation function. This enhancement allows the model to better
utilize features from both shallow and deep convolutional layers, improving semantic representation
and adaptability to complex tasks. A comprehensive ALD dataset, AppleLeaf9, was created by
combining four different datasets, including images from wild environments. This dataset contains a
variety of ALDs and healthy leaves, with 94% of the images taken in the wild.The EfficientNet-MG
model was trained and tested on the AppleLeaf9 dataset, achieving an accuracy of 99.11% and
requiring fewer parameters (8.42 million) compared to five classical CNN models. This demonstrates
the model's efficiency and robustness for ALD identification in the wild.[28] The study employed
transfer learning to fine-tune pre-trained models and used data augmentation techniques to prevent
overfitting. The evaluation is conducted on the publicly available Plant Village dataset, which
includes 38 classes of plant leaf diseases from 14 crops. proposed DENN achieves high classification
accuracy on the Plant Village dataset, outperforming individual pre-trained models, the DENN with
transfer learning is a robust and efficient approach for plant leaf disease classification, offering high
accuracy with minimal computational resources.[29] The review discussed various ML and DL
techniques used for plant disease detection, including Naive Bayes, Decision Tree, K-Nearest
Neighbour, Support Vector Machine, Random Forest, Multilayer Perceptron, and Logistic Regression.
It also explores deep learning models like Convolutional Neural Networks (CNN), VGG-16, VGG-19,
and Inception-V4, highlighting their effectiveness in image-based disease detection.[30] The proposed
LTriTP method involved several steps: image preprocessing, feature extraction using the LTriTP
descriptor, and classification. The method's performance is evaluated using the PlantVillage dataset,
which includes images of tomato leaves with six different diseases. The LTriTP method achieved high
classification accuracy across multiple disease classes, with the best performance observed using the
Cubic SVM classifier (97.80%).

References

[1] Wang Linfeng Liu Jiayao Liu Yong Wang Yunsheng Xu Shipu “A lightweight tomato leaf disease
identification method based on shared-twin neural networks” IET Image Process. 2024;1–13. DOI:
10.1049/ipr2.13094.

[2] Sunil C.K. , Jaidhar C.D., Nagamma Patil “Tomato plant disease classification using Multilevel
Feature Fusion with adaptive channel spatial and pixel attention mechanism”
https://doi.org/10.1016/j.eswa.2023.120381

[3] Neeraj S Kumara , Jyothika Sonya , Arathi Premkumara , Meenakshi Ra , Jyothisha J Naira
“Transfer Learning-based Object Detection Models for Improved Diagnosis of Tomato Leaf Disease”
10.1016/j.procs.2024.04.286.

[4] Showmick Guha Paul a , Al Amin Biswas b , Arpa Saha a , Md. Sabab Zulfiker a , Nadia Afrin
Ritu c , Ifrat Zahan a , Mushfiqur Rahman a , Mohammad Ashraful Islam “A real-time application-
based convolutional neural network approach for tomato leaf disease classification”
https://doi.org/10.1016/j.array.2023.100313.

[5] Shengyi Zhao, Yun Peng, Jizhan Liu and Shuo Wu “Tomato Leaf Disease Diagnosis Based on
Improved Convolution Neural Network by Attention Module” Agriculture 2021, 11, 651.
https://doi.org/10.3390/agriculture11070651

[6] Wenbo Liu *, Yongsen Zhai and Yu Xia “Tomato Leaf Disease Identification Method Based on
Improved YOLOX” Agronomy 2023, 13, 1455. https://doi.org/10.3390/agronomy13061455

[7] Omneya Attallah “Tomato Leaf Disease Classification via Compact Convolutional Neural
Networks with Transfer Learning and Feature Selection” Horticulturae 2023, 9, 149.
https://doi.org/10.3390/horticulturae9020149.

[8] Harsh Baheti, Anuradha Thakare, Yash Bhople, Sudarshan Darekar, Om Dodmani “Machine
Learning Algorithm for Detection And Classification of Tomato Plant Leaf Disease” 978-1-6654-
2168-3/22/$31.00 ©2022 IEEE | DOI: 10.1109/I2CT54291.2022.9824029

[9] Rajasekaran Thangaraj S. Anandamurugan Vishnu Kumar Kaliappan “Automated tomato leaf
disease classifcation using transfer learning-based deep convolution neural network”
https://doi.org/10.1007/s41348-020-00403-0

[10] Lerina Aversano, Mario Luca Bernardi, Marta Cimitile, Martina Iammarino, Stefano Rondinella
“Tomato diseases Classification Based on VGG and Transfer Learning” 978-1-7281-8783-
9/20/$31.00 ©2020 IEEE.
[11] Vemakoti Santosh Krishna, Datti Rakesh, Sachikanta Dash, Bidush Kumar Sahoo, Sasmita
padhy, Mamata Nayak “Tomato Leaf Disease Detection using Neural Networks” 978-1-6654-5493-
3/22/$31.00 ©2022 IEEE.

[12] Amine Mezenner, Hassiba Nemmour, Youcef Chibani, Adel Hafiane “Tomato Plant Leaf Disease
Classification based on CNN features and Support Vector Machines” 978-1-6654-1741-9/22/$31.00
©2022 IEEE.

[13] Thanjai Vadivel and R. Suguna “Automatic recognition of tomato leaf disease using fast
enhanced learning with image processing” ACTA AGRICULTURAE SCANDINAVICA, SECTION B
— SOIL & PLANT SCIENCE 2022, VOL. 72, NO. 1, 312–324
https://doi.org/10.1080/09064710.2021.1976266.

[14] Jie Hang , Dexiang Zhang , Peng Chen , Jun Zhang and Bing Wang “Classification of Plant Leaf
Diseases Based on Improved Convolutional Neural Network” Sensors 2019, 19, 4161;
doi:10.3390/s19194161.

[15]NareshV. ,YogeswararaoG.,NageswararaoNaikB.,MalmathanrajR., and PalanisamyP “Empirical

[16] Dasen Li, Zhendong Yin , Yanlong Zhao, Jiqing Li, Hongjun Zhang “Rehearsal-based class-
incremental learning approaches for plant disease classification” 0168-1699/© 2024
https://doi.org/10.1016/j.compag.2024.109211

[17] Husna *, Ahamed , Jannat , Muhit , Islam “Characterization of Potato Leaf Disease by Digital
Image Processing Technique” Vol 5No 2June2024Pages 13-18e-ISSN 2708-5694J. Agric. Food
Environ. 5(2): 13-18, 2024.

[18] Vibhor Kumar Vishnoi, Krishan Kumar, Brajesh Kumar , Shashank Mohan ,Arfat Ahmad Khan
“Detection of Apple Plant Diseases Using Leaf Images Through Convolutional Neural Network”
volume 11, 2023 doi: 10.1109/ACCESS.2022.3232917.

[19] Bin Wang, HuaYang, Shujuan Zhang and Lili Li “Identification of Multiple Diseases in Apple
Leaf Based on Optimized Lightweight Convolutional Neural Network.” Plants 2024, 13, 1535. https://
doi.org/10.3390/plants13111535 .

[20] Abdelmalik Ouamane, Ammar Chouchane, Yassine Himeur, Abderrazak Debilou,Abbesamira,

[21] Yun Zhao, Chengqiang Lin, Na Wu and Xing X “ APEIOU Integration for Enhanced
YOLOV7: Achieving Efficient Plant Disease Detection” Agriculture 2024, 14, 820.
https://doi.org/10.3390/agriculture14060820 .

[22] Weisi Dai, Wenke Zhu, Guoxiong Zhou, Genhua Liu, Jiaxin Xu, Hongliang Zhou, Yahui Hu,
Zewei Liu, Jinyang Li, Liujun Li “AISOA-SSformer: An effective image segmentation method for
rice leaf disease based on the Transformer architecture” https://spj.science.org on July 06, 2024.

[23] H. Sabrol and K. Satish “Tomato Plant Disease Classification in Digital Images using
Classification Tree” 978-1-5090-0396-9/16/$31.00 ©2016 IEEE.
[24] Mohit Agarwala, Abhishek Singhb, Siddhartha Arjariac , Amit Sinhad, Suneet Guptaa “ToLeD:
Tomato Leaf Disease Detection using Convolution Neural Network” Procedia Computer Science 167
(2020) 293–301
[25] Alvaro Fuentes, Sook Yoon , Sang Cheol Kim and Dong Sun Park “ARobust Deep-Learning-
Based Detector for Real-Time Tomato Plant Diseases and Pests Recognition” Sensors 2017, 17, 2022;
doi:10.3390/s17092022.
[26] Konstantinos P. Ferentinos “Deep learning models for plant disease detection and diagnosis”
https://doi.org/10.1016/j.compag.2018.01.009
[27] Qing Yang, Shukai Duan and Lidan Wang “Efficient Identification of Apple Leaf Diseases in the
Wild Using Convolutional Neural Networks” Agronomy 2022, 12, 2784.
https://doi.org/10.3390/agronomy12112784
[28] Sasikala Vallabhajosyula, Venkatramaphanikumar Sistla, Venkata Krishna Kishore Kolli
“Transfer learning-based deep ensemble neural network for plant leaf disease detection”
https://doi.org/10.1007/s41348-021-00465-8
[29] C Jackulin , S. Murugavalli “A comprehensive review on detection of plant disease using
machine learning and deep learning approaches” Measurement: Sensors 24 (2022) 100441
https://doi.org/10.1016/j.measen.2022.100441
[30] Wakeel Ahmad, Syed M. Adnan, Aun Irtaza “Local triangular-ternary pattern: a novel feature
descriptor for plant leaf disease detection” Multimedia Tools and Applications (2024) 83:20215–
20241 https://doi.org/10.1007/s11042-023-16420-8