A Comparative Survey on Paddy Leaf Diseases

Detection using Convolutional Neural Network

Algorithms
2023 Second International Conference on Advances in Computational Intelligence and Communication (ICACIC) | 979-8-3503-1845-6/23/$31.00 ©2023 IEEE | DOI: 10.1109/ICACIC59454.2023.10435302

Dasa Hemanth Kumar Samundiswary Punniakodi

Department of Electronics Engineering Department of Electronics Engineering
Pondicherry University Pondicherry University
Puducherry, India Puducherry, India
[email protected] [email protected]

Abstract—Despite recent difficulties in rice cultivation, rice in Section II. Various rice leaf diseases are listed in Section
leaf diseases are still identified using traditional methods. This III. Subsequently, a performance evaluation comparison is
leads to a fall in production of the crops. Initial detection allows presented in Section IV, the main issues are covered in
farmers to increase output rates and efficiently take the Section V, and the document is concluded in Section VI.
necessary treatment measures. Convolutional Neural Network
(CNN) algorithms' quick development has significantly II. RELATED WORKS
modified agricultural practices, particularly in disease
identification. This paper comprehensively examines the use of Yibin Wang et al. proposed an Attention-based
CNN-based approaches for identifying illnesses in rice leaves. Depthwise Separable Neural Network with Bayesian
Given the rising need for rice production and quality on a global Optimisation (ADSNN-BO) by considering a MobileNet
scale, accurate and quick disease diagnosis is essential. This structure to identify and categorise rice diseases from the
review study also deals with the different CNN architectures and
the corresponding modifications to detect rice disease. It also photos of rice leaves. The Bayesian optimisation technique is
includes single- and multi-class classification algorithms. In also used to adjust the hyperparameters. Four categories for
addition, examining important datasets, preprocessing methods, rice disease are included in a publicly available dataset that
and augmentation strategies reveals their impact on the serves as the basis for cross-validated classification
performance of the developed models. Comparative evaluations experiments. Increases in the accuracy of training
of the surveyed papers highlight the potential and constraints of
various CNN models and suggest fresh directions for
experimental outcomes suggest that the mobile-friendly
investigations. This study highlights the potential of CNNs to ADSNN-BO model performs well among all modern and
completely reinvent across automated and effective verification, current models by obtaining a test rate of 94.65%. The
which ultimately increases agricultural productivity. suggested model's interpretability in feature analysis is tested
Keywords—Paddy Leaf Disease, Convolutional Neural using filter visualisation and activation maps approaches.
Network, Faster R-CNN, Deep Convolutional Neural Network. Results show that the proposed attention-based technique may
successfully instruct the ADSNN-BO technique to learn
I. INTRODUCTION essential features. This study's results will help validate the
application of artificial intelligence for agricultural purposes
Rice is a widely consumed, nutritious food produced in
for the speedy recognition and control of leaf diseases[1].
large quantities. Today, rice is more farmed than any other
notable grain because it is still a staple meal and grain for Bifta Sama Bari et al. developed a Faster Region-based
humans. Rice diseases continue to be produced in Convolutional Neural Network (Faster R-CNN) to recognise
conventional ways despite the growing difficulty of farming rice leaf diseases. The Faster R-CNN approach uses an
rice[1-3]. Early detection assists farmers in treating the enhanced RPN architecture to produce potential areas. To
problem correctly and further increasing the yield [4,5] increase the resilience of the Faster R-CNN technique, the
effectively. Visual inspection, which examines leaves for network is trained using openly available web datasets with
disease symptoms, including discolouration, wilting, or the
real-world datasets of rice leaves. The suggested approach
presence of disease-causing organisms, is one of the frequent
techniques for identifying leaf diseases [6-8]. While this effectively automatically diagnosed three distinct diseases of
approach can help to diagnose some illnesses, it can be time- rice leaves, including brown spot, rice blast, and hispa, at a
consuming and may miss infections without apparent precision of 98.09%, 98.85%, and 99.17%, respectively.
symptoms. Thus, visual inspection is a standard method, but Furthermore, the model provides an accuracy of 99.25% in
novel technologies like imaging and sensor systems are determining a healthy rice leaf. The outcomes obtained here
promising techniques to diagnose illnesses earlier and more showed that the Faster R-CNN model delivers the improved
precisely[9,10]. Early diagnosis of crop diseases is essential performance rice leaf illness identification method that could
to minimize agricultural losses and guarantee food security more accurately and quickly identify the most widespread rice
[11-13]. CNNs have become powerful tools for automated illnesses[2].
disease diagnosis in leaves with advanced deep-learning
methodologies [14-17]. This paper includes the study of Krishnamoorthy N et al. proposed that a specific nature
various CNN-based rice disease detection techniques. The of the CNN model called InceptionResNetV2 utilises transfer
remaining part of the paper is organised as follows: Related learning technique to identify infections in photos of rice
works and techniques used in this investigation are outlined

2023 Second International Conference on Advances in Computational Intelligence and Communication

(ICACIC), 7th & 8th Dec.2023, Puducherry Technological University, India ©2023 IEEE

Authorized licensed use limited to: Manipal University. Downloaded on February 27,2024 at 17:18:50 UTC from IEEE Xplore. Restrictions apply.
leaves. The recommended model's parameters were ready for accuracy frameworks based on the Multi-Layer Perceptron
classification, providing a respectable accuracy of 95.67[3]. (MLP) and CNN architectures, 91.25% and 82.03%,
respectively. The Rice-Fusion paradigm for multimodal data
Rahul Sharma et al. suggested a CNN model to classify
fusion is recommended, which surpasses the outcomes of the
leaf diseases in potato and rice plants. Biological blight, blast,
findings of the experiments, unimodal frameworks
brown spot, and tungro diseases have been identified in rice
analysis[7].
leaves. The division of potato leaves into three categories:
diseased leaves, early blight, and late blight. This analysis To assess biotic stress in rice fields with the
uses a dataset of 5932 images of rice leaves and 1,500 images computational power of portable devices, Panjal Joshi et al.
of potato leaves. The suggested model can extract hidden suggested RiceBioS, an AI-based DL-enabled handheld
patterns in the raw photos and categorise images of rice and device. RiceBioS uses Edge-as-a-Service (EaaS) to divide
potato leaves with 99.58 and 97.66 accuracy, respectively. rice crop photos into stressed and healthy groups. The biotic
The outcomes show that the suggested CNN model stress state is further divided into two distinct infection types:
outperforms all other ML classification algorithms, including bacterial rice leaf blight and fungal rice blast by pruning the
K-Nearest Neighbours (KNN), Support Vector Machine shrunken deep learning classification model and
(SVM), and Random Forest and Decision Tree[4]. incorporating feature extraction workflow, an automated RoI
recognition, adaptive thresholding, utilising hierarchical
It focuses primarily on three common diseases that harm
masking methods for dimensionality reduction. Although
rice: brown spot, leaf blight, and leaf smut caused by fungus
RiceBioS has a testing accuracy of 93.25%, its after-
and bacteria, respectively. According to Santosh Kumar
deployment performance on a smartphone is barely
Upadhyay et al., an effective way to diagnose and recognise
noticeable. This state-of-the-art technology aids farmers in
rice leaf disease is developed due to the leaf's size, shape, and
making knowledgeable judgements based on current
colour images of lesions. This model's picture banalisation
information offered by RiceBioS's simple-to-use mobile
process employs Otsu's method of global thresholding to
application interface[8].
eliminate image background noise. To determine the three
rice diseases, this model is built around fully connected CNN, P.S. Thakur et al. proposed the lightweight convolutional
which is trained with 4000 images of each diseased leaf and neural network Visual Geometry Group-InceptionV7
4000 images of healthy rice leaves. It is verified through this Convolutional Neural Network (VGG-ICNN) to identify crop
method that the fully linked CNN approach achieves a 99.7% illnesses from plant-leaf photos. The VGG-ICNN has about 6
accuracy across the dataset[5]. million fewer features than most other high-performing DL
models currently on the market. Many crop varieties are
SK Mahmudul Hasan et al. have suggested a
represented in five separate public datasets that assess the
revolutionary using a deep learning model, the foundation of
model's performance. These include datasets for several crop
the residual connection and the inception layer. The number
species, such as Embrapa and PlantVillage, and databases for
of features is reduced by using depth-wise separable
just one crop, such as apples, rice and maize, with four and
convolution. The proposed model was trained and tested on
five categories and 93 and 38 classes, respectively. Using the
three different datasets related to plant diseases. The
PlantVillage dataset, the results of experiments demonstrate
performance accuracy for the plantvillage dataset is 99.39,
that the system outperforms various present approaches for
and 99.66 for the rice illness is 76.59 on the cassava dataset.
crop disease identification. Compared to several recent
The proposed model provides more accurate results with
lightweight CNN models, the VGG-ICNN model consistently
fewer parameters when compared to state-of-the-art deep
outperforms the other models on all five datasets[9].
learning methods[6].
J. Chen et al. proposed that the MobInc-Net valid
Rutuja R. Patil et al. proposed Rice-Fusion, a cutting-
lightweight network architecture recognizes and
edge multimodal data fusion system to detect rice disease. It
identifies crop diseases. The Inception module was improved
is influenced by the utility of CNN's approaches and IoT for
using depth and point-wise convolutions instead of the
agriculture. A comprehensive and reliable illness diagnosis
original ones. Modified Inception module(M-Inception)
for rice disease requires the fusion of heterogeneous
combined using the backbone extractor, the pre-trained
modalities because a single modality-based diagnosis may
MobileNet, to extract superior picture characteristics.
not be valid. This offers a novel viewpoint in the realm of rice
Following that, the Single Shot multi-box Detector(SSD)
disease diagnosis. They manually gathered the dataset using
block and the relationship between the Softmax layer and the
3200 rice health class samples, a camera, and agro-
number of classes separately inserted underlying the support
meteorological sensors. The agro-meteorological data
network for categorising and detecting different types of crop
collected from sensors is extracted using the rice-fusion
disease. Two-step transfer learning is used when training
framework into numerical features. It then removes the visual
models to create an effective model. According to
elements from the photos of the rice. To further combine
experimental results, the proposed technique achieved the
these extracted features, a concatenation layer is utilised next
necessary efficiency, 97.89 for the local dataset and an
to a dense layer, which provides Rice-Fusion testing accuracy
average detection rate of 99.21 on the general dataset[10].
of 95.31%, more significant than the alternative unimodal

Authorized licensed use limited to: Manipal University. Downloaded on February 27,2024 at 17:18:50 UTC from IEEE Xplore. Restrictions apply.
 G. Latif et al. suggested a transfer-based DCNN, short dilated convolution kernels combined with standard CNN
for Deep Convolutional Neural Network, a technique for convolution kernels.
precisely detecting and classifying rice leaf disease. The
revised proposed methodology includes a modified transfer Dilated convolution offers the benefits of lower computing
learning algorithm based on VGG19. A total of six classes— costs and less memory use, while GAP prevents overfitting.
leaf scald, bacterial leaf blight, brown spot, healthy, leaf With CNN, these two novel ideas are put into practice.
blast, and narrow brown spot—can be reliably diagnosed Performance indicators, including recall, precision, f1-score,
using the suggested revised approach. Using the non- and accuracy, are used to compare the outcomes of this
normalized enhanced dataset, the maximum average technique with those of other hybrid learning techniques. The
accuracy is 96.08%. Precision, specificity, recall, and F1- proposed model classified diseases into four categories: blast,
score were correspondingly 0.9620, 0.9617, 0.9921, and bacterial blight, tungro, and brown spot. Performance data
0.9616 compared to the same dataset with similar reveal that, compared to the traditional CNN model, the
dimensions, using similar methods—the proposed modified DCNN model with GAP increases training accuracy at an
technique provided significantly superior results[11]. average of 5.49% in the same experimental configuration. The
outcomes are contrasted with conventional and hybrid
R.R. Patil and Kumar suggested a rice transformer
learning techniques[14].
combining inputs from field-based picture data and
agricultural sensors. The two components of the proposed B. Jangid and Sharma et al.'s research revolves around
system are the sensor and image branches. In particular, the VGG-16 DL model, which is used as the main component
characteristics from both modalities are obtained with the of the methodology, and a Flask web application is used for
attention technique. The extracted parameters are later user interaction. This methodology's essential step is Flask
transmitted in a crisscross pattern as input referring to the Web Application: The researchers created a Flask-based web
cross-attention unit to enhance recognition of the traits application to make the model available to users. Using this
unique to rice diseases. To accurately classify the rice tool, users can upload images of rice plants to check for
diseases, the retrieved features are combined, pooled, and diseases[15].
sent via the Softmax classifier. The dataset was compiled
Md. S. H. Shovon et al. introduced PlantDet as a brand-
using 4200 real-time samples from rice farms and is a
new multi-model ensemble approach based on deep learning
customized dataset. The suggested Rice transformer model
for identifying plant diseases. The suggested strategy uses the
yields an accuracy of around 97.38% for preventing rice
advantages of numerous models to overcome the weaknesses
disease, which is somewhat encouraging[12].
of a single model, improving accuracy and resilience. The
O. Debnath and Saha’s study provides an intelligent experimental findings demonstrate that PlantDet can correctly
farming strategy utilising the IoT and machine learning. The recognise leaf diseases in a different dataset. Early and precise
uniqueness of this model is that the first-ever use of diagnosis of plant diseases made possible by this research
convolutional neural networks for the rice paddy leaves with positively contributes to agricultural technology. It helps to
brown spot disease is quickly identified by these symptoms. increase crop output and quality[16].
Neural networks are used in Deep Learning to implement AI.
Syed Rehan Shah et al. offer a practical comparative
In this study, real-time data-based image recognition and
analysis of five deep learning architectures such as Inception
preprocessing are applied. Feature extraction and data
V3, VGG 19, VGG 16, CNN, and ResNet50, for the early
preparation have been performed using a self-created image-
diagnosis of a rice disease: Syed Rehan Shah et al. findings
processing program. Personally collected data for training
emphasise the importance of selecting the exemplary
and testing from rice fields were utilised to create the Tensor
architecture for precisely detecting diseases. Out of these five
flow with the Keras framework. The model’s accuracy is
techniques, ResNet50 is showing the best results. The results
about 97.701%, offering the potential to limit the overall
help researchers to choose the best model for comparable
production losses on a national and international level[13].
agricultural applications[17]
S. Senthil Pandi et al. addressed the challenges of
applying advanced imaging and Deep Learning(DL)
techniques for plant leaf disease detection.
Conventional CNN has the problem of consuming a great
deal of processing resources. To address this problem, a
Dilated Convolutional Neural Network (DCNN) model with
Global Average Pooling (GAP) is employed. The wholly
connected layer in this model is swapped out for a
conventional CNN with Global Average Pooling, and the

Authorized licensed use limited to: Manipal University. Downloaded on February 27,2024 at 17:18:50 UTC from IEEE Xplore. Restrictions apply.
III. COMPARATIVE ANALYSIS OF VARIOUS CNN-BASED LEAF
DISEASE IDENTIFICATION FOR PADDY CROPS

TABLE I. TECHNIQUES AND DISEASE IDENTIFICATION OF Leaf Blight, Narrow

PADDY CROPS brown spot, Leaf blast
June- CNN-
Refe Month Accurac [17] Leaf blast 99.69%
2023 ResNet50
renc and Algorithm Diseases Identified y
e Year
March-
rice hispa damage, rice The comparative study of the above table concludes that
[1] ADSNNBO leaf blast, and brown 94.65
2021 accuracy achieved about 99.69 but identified a single disease.
spot
Faster R- 98.09,
[2]
April-
CNN
Hispa, rice blast,
98.85,
Performance Comparison
2021 and brown spot
framework 99.17
Performance evaluation of related works such as Accuracy of
Leaf Blast(LB), Brown
May- InceptionRes the disease detection, Precision and F1-Score are shown in the
[3] Spot(BS), Healthy, and 95.67
2021 NetV2
Bacterial Blight(BB) Figures 1, 2 and 3 below:
CNN Rice diseases: blast,
Classification tungro, bacterial blight, Rice-
: SVM, KNN, and brown spot ACCURACY
July- 99.58,
[4] Decision diseases
2021 Potato- 105
Tree, Potato diseases:
Random healthy leaves and 97.66 100
Forest early blight late blight
95
Octobe Bacterial Leaf Blight, 99.1 &
[5] DCNN 90
r-2021 Brown Spot, Leaf smut 99.7
Inception 85
Januar module and Bacterial blight, Brown
[6] 99.66 80
y-2022 Residual spot, Blast, Tungro

[9]
[8]
[7]
[6]
[5]
[4]
[3]
[2]
[1]
[17]
[16]
[15]
[14]
[13]
[12]
[11]
[10]
Network
Multimodal
Early Fusion ACCURACY
method Brown spot, Rice blast,
Januar
[7] Unimodal healthy, and Bacterial 95.31 Fig.1. Accuracy vs Various CNN approaches
y-2022
CNN and blight
MLP
architectures
March- CNN- rice blast and bacterial PRECISION
[8] 93.25
2022 RiseBioS leaf blight
Bacterial Leaf 150
Streak(BLS), White
June-
[9] VGG-ICNN Tip, Leaf Scald(LS), 96.67 100
2022
Leaf Smut, and
Stackburn 50
Sheath blight, Rice
blast, Brown spot, Leaf
scald, leaf smut, Rice 0
[17] [16] [15] [14] [13] [12] [11] [10] [9] [8] [7] [6] [5] [4] [3] [2] [1]
stackburn, Rice false
July- smut, White tip, Rice PRECISION
[10] MobInc-Net 97.89
2022 stem rot(RSR), Rice
sheath rot(RSR),
Fig.2. Precision vs Various CNN approaches
Bacterial leaf
blight(BLB), Rice
stripe blight
leaf scald, narrow F1-SCORE
brown spot, leaf blast,
August 150
[11] VGG-19 bacterial leaf blight, 96.08
-2022
brown spot, and 100
healthy
Deep 50
Rice Blast(RB), Brown
August Convolution
[12] Spot(BS), Healthy, and 96.9
-2022 Neuro-Fuzzy 0
Bacterial Blight(BB)
[9]
[8]
[7]
[6]
[5]
[4]
[3]
[2]
[1]
[17]
[16]
[15]
[14]
[13]
[12]
[11]
[10]

Network
August
[13] CNN Brown spot 97.701
-2022 F1-SCORE
DCNN and
Octobe Global bacterial blight, blast,
[14] 96.5 Fig.3. F1-Score vs CNN approaches
r-2022 Average brown spot and turgo
Pooling
healthy, leaf blast, IV. LIMITATIONS
March- CNN-VGG-
[15] brown spot, neck 89.38
2023 16 There are several limitations observed in the literature
blast
[16]
April-
PlantDet
Leaf scald, Healthy,
98.53% review. Training time for the models should be decreased as
2023 Brown spot, Bacterial much as possible. Experimental setups took more epochs to

Authorized licensed use limited to: Manipal University. Downloaded on February 27,2024 at 17:18:50 UTC from IEEE Xplore. Restrictions apply.
identify diseases accurately. Most of the studies utilised free [13] O. Debnath and H. N. Saha, “An IoT-based intelligent farming using
CNN for early disease detection in rice paddy,” Microprocessors and
resources (Google Colab & Jupyter Notebook). Microsystems, Elsevier, vol. 94, p. 104631, Oct. 2022, doi:
10.1016/j.micpro.2022.104631.
[14] S. Senthil Pandi, A. Senthilselvi, J. Gitanjali, K. ArivuSelvan, J.
V. CONCLUSION Gopal, and J. Vellingiri, “Rice plant disease classification using a
dilated convolutional neural network with global average pooling,”
CNN methods often detect diseases that contaminate rice Ecological Modelling, Elsevier, vol. 474, p. 110166, Dec. 2022, doi:
leaves. This paper rigorously examines the available CNN- 10.1016/j.ecolmodel.2022.110166.
[15] B. Jangid and R. S. Sharma, “Rice Disease Detection Using Deep
based approaches for leaf disease identification in paddy Learning VGG-16 Model and Flask,” International Journal of
crops. Further, the performance of well-known CNN Scientific Research in Engineering and Management (IJSREM),
frameworks is compared and analysed. It is also noticed that vol:07, Issue: 03, 2023. doi: 10.55041/IJSREM17874.
better accuracy performance is obtained using pre-trained [16] Md. S. H. Shovon, S. J. Mozumder, O. K. Pal, M. F. Mridha, N. Asai,
and J. Shin, “PlantDet: A Robust Multi-Model Ensemble Method
models rather than creating them from scratch using rice leaf Based on Deep Learning For Plant Disease Detection,” IEEE Access,
datasets with the aid of CNN approaches. However, creating vol. 11, pp. 34846–34859, 2023, doi:
10.1109/ACCESS.2023.3264835.
highly effective detection methods using vast datasets of
[17] Syed Rehan Shah, S. Qadri, H. Bibi, S. M. W. Shah, M. I. Sharif, and
many leaf diseases will be a crucial future influence. By F. Marinello, “Comparing Inception V3, VGG 16, VGG 19, CNN, and
necessitating sizable generalised datasets, this would also ResNet 50: A Case Study on Early Detection of a Rice Disease,”
alleviate the class imbalance. Agronomy, MDPI, vol. 13, Issue 6, p1633, June 2023, doi
10.3390/agronomy13061633.
REFERENCES
[1] Yibin Wang, H. Wang, and Z. Peng, “Rice diseases detection and
classification using attention-based neural network and Bayesian
optimization,” Expert Systems with Applications, Elsevier, vol. 178,
p. 114770, Sep. 2021, doi: 10.1016/j.eswa.2021.114770.
[2] Bifta Sama Bari, Md Nahidul Islam, Mamunur Rashid, Md Jahid
Hasan, Mohd Azraai Mohd Ramzan, Rabiu Muazu Musa, Ahmad
Fakhri Ab Nasir, and Anwar P.P. Abdul Majeed, “A real-time
approach of diagnosing rice leaf disease using deep learning-based
faster R-CNN framework,” PeerJ Computer Science, vol. 7, p. e432,
Apr. 2021, doi: 10.7717/peerj-cs.432.
[3] Krishnamoorthy N, L. V. Narasimha Prasad, C. S. Pavan Kumar, B.
Subedi, H. B. Abraha, and S. V e, “Rice leaf diseases prediction using
deep neural networks with transfer learning,” Environmental
Research, Elsevier, vol. 198, p. 111275, Jul. 2021, doi:
10.1016/j.envres.2021.111275.
[4] Rahul Sharma, Amar Singh, Kavita, N. Z. Jhanjhi, Mehedi Masud,
Emad Sami Jaha5 and Sahil Verma, “Plant Disease Diagnosis and
Image Classification Using Deep Learning,” Computers, Materials &
Continua, Tech Science Press, vol. 71, pp. 2125–2140, Dec. 2021, doi:
10.32604/cmc.2022.020017.
[5] Santosh Kumar Upadhyay and A. Kumar, “A novel approach for rice
plant diseases classification with deep convolutional neural network,”
International Journal of Information Technology, vol. 14, no. 1, pp.
185–199, Feb. 2022, doi: 10.1007/s41870-021-00817-5.
[6] SK Mahmudul Hasan and A. K. Maji, “Plant Disease Identification
Using a Novel Convolutional Neural Network,” IEEE Access, vol. 10,
pp. 5390–5401, 2022, doi: 10.1109/ACCESS.2022.3141371.
[7] Rutuja R. Patil and S. Kumar, “Rice-Fusion: A Multimodality Data
Fusion Framework for Rice Disease Diagnosis,” IEEE Access, vol.
10, pp. 5207–5222, 2022, doi: 10.1109/ACCESS.2022.3140815.
[8] Panjal Joshi, D. Das, V. Udutalapally, M. K. Pradhan, and S. Misra,
“RiceBioS: Identification of Biotic Stress in Rice Crops Using Edge-
as-a-Service,” IEEE Sensors Journal, vol. 22, no. 5, pp. 4616–4624,
Mar. 2022, doi: 10.1109/JSEN.2022.3143950.
[9] P. S. Thakur, T. Sheorey, and A. Ojha, “VGG-ICNN: A Lightweight
CNN model for crop disease identification,” Multimedia Tools and
Applications, vol. 82, no. 1, pp. 497–520, Jan. 2023, doi:
10.1007/s11042-022-13144-z.
[10] J. Chen, W. Chen, A. Zeb, S. Yang, and D. Zhang, “Lightweight
Inception Networks for the Recognition and Detection of Rice Plant
Diseases,” IEEE Sensors Journal, vol. 22, no. 14, pp. 14628–14638,
Jul. 2022, doi: 10.1109/JSEN.2022.3182304.
[11] G. Latif, S. E. Abdelhamid, R. E. Mallouhy, J. Alghazo, and Z. A.
Kazimi, “Deep Learning Utilization in Agriculture: Detection of Rice
Plant Diseases Using an Improved CNN Model,” Plants, MDPI, vol.
11, no. 17, Art. no. 17, Jan. 2022, doi: 10.3390/plants11172230.
[12] R. R. Patil and S. Kumar, “Rice Transformer: A Novel Integrated
Management System for Controlling Rice Diseases,” IEEE Access,
vol. 10, pp. 87698–87714, 2022, doi:
10.1109/ACCESS.2022.3200688.

Authorized licensed use limited to: Manipal University. Downloaded on February 27,2024 at 17:18:50 UTC from IEEE Xplore. Restrictions apply.