SlideShare a Scribd company logo
1 of 71
Presented by
D N Kiran Pandiri
Reg. No. 19-3-04-118
Supervisor
Dr. R Murugan,
Assistant Professor
Department of Electronics and Communication Engineering
National Institute of Technology, Silchar
Soil and Crop Leaf Disease Classification using
Artificial Intelligence
Proposed Tentative Title
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 2
I Welcome the Doctoral Committee to
the Enhancement Seminar
 Chairman : Dr. T Khan, Asst Prof, ECE
 External Examiner : Dr. Sumantra Dutta Roy, Professor, EE, IIT Delhi
 Member : Dr. Chandrajit Choudhury, Asst Prof, ECE
 Nominee of Chairman, Senate : Dr. Asha Rani M A, Asst Prof, EE
 Supervisor : Dr. R Murugan, Asst Prof, ECE
Outline of the presentation
 Broad area of research
 Introduction
 Literature survey
 Problem statements
 Objectives
 Conclusion
 Publications
 Works under communication
 References
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 3
Broad area of research
 Image Processing in Agriculture
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 4
Introduction
 It is a practice of cultivating
plants and livestock.
 It has been practicing from
many centuries by our
ancestors and by us till now.
 Agriculture with its allied
sectors is the largest source
of livelihood in many
countries across the globe.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 5
Fig.1. Image representation of cultivation and livestocks. [1]
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 6
The agriculture and forest lands decreased for the past two decades [2][3].
The global food demand has been increasing with increase in population.
Fig. 2. Agriculture and forest land area declined in statistics from 2000 – 2018 [2]
Crop Loss due to diseases
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 7
0
50
100
Yield
Loss
Between 20 - 40 percent of Global crop
yields are reduced each year due to the
damage wrought by plant pests and
diseases.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 8
Usage of pesticides and fertilizers
Fig. 3. Percentage increase in the usage of fertilizers [2]
Decrease in work force
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 9
Fig 4. The decline of workforce in the last two decades [2]
Digital Transformation of Agriculture
 There is a need for optimizing the agriculture practices without putting extra
environmental burden.
 The digital transformation of the agriculture sector is to meet the increasing food
demand of the global population explosion without affecting the environmental
conditions and maintaining the profit to the farmers.
 Digitalization of agriculture can reduce the burden of employee workforce, crop
management, soil management and many other.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 10
Motivation
 Digitalization of the agriculture sector that reduces the use of fertilizers and pesticides
and increases the crop yield by maintaining profits to the farmers.
 We have consider two key components which can reduce the usage of fertilizers and
pesticides in increasing the crop yield is
1. Soil type classification
2. Plant leaf disease classification
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 11
Soil Type Classification
 Soil is the primary component in agriculture.
 Soil type identification is the key to the success of site-specific crop management.
 Soil management helps in varying the tillage [4].
 Identifying the soil type can reduce or minimize the soil tillage by reducing the
production cost.
 Based on the soil type, the soil conditions like soil texture and moisture content can
be adjusted, which are important for seed and fertilizer placement.
 There is no effective research on physical properties of the soil like soil type, porosity,
texture, structure, density, etc.,
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 12
 The conventional methods used for identifying the soil type are
 Laboratory testing [7][8]
 In-situ testing [5][6]
 All these methods are time-consuming, high cost, and require proficiency to analyze
soil type.
 By using digital techniques, researchers used texture, color, particle size, pattern, pH
level, and organic matter as features in distinguishing the different types of soils.
 The soil type is identified based on the texture of the soil sample.
 The soil texture varies on the percentage of sand, silt, and clay present in the soil
sample.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 13
Sample Soil Images
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 14
Fig 5. Sample Soil images. (a) Loamy sand, (b)&(d) sand, (c),(e)&(f) clay, and (g) sandy loam.
(a) (b) (c) (d)
(e) (f) (g)
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 15
[Ref]
Year
Title Input Data Methods Focus
[9]2015
Soil type classification and mapping using
hyperspectral remote sensing data.
Hyperspectral
image
Gaussian SVM
Classify the soils based on the colour with
hyperspectral data
[10]2019
Compressive spectral imaging system for soil
classification with three-dimensional
convolutional neural network.
3D CNN
Hyperspectral images are used in
classification of soil type. PCA is used in
reducing the feature dimensions.
[11]2016
A smartphone-based soil color sensor: For soil
type classification.
Smart Phone
Camera with
spectral lenses
Linear Discriminant
Analysis
Classified the soil types based on the
Munsel color. The color of the soil
identified by using the spectral
wavelength.
[12]2020
Classification of soil aggregates: A novel approach
based on deep learning.
Stereo pair
images
CNN
Deep Learning methods in classifying
aggregate soil
[13]2012
Soil texture classification algorithm using RGB
characteristics of soil images.
CCD Camera
Images
Linear Regression
Using RGB histogram of the CCD camera
images, classified the soil types
Literature Review on Soil Classification
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 16
[Ref]
Year
Title Input Data Methods Focus
[14]2018
Soil Characterization and Classification: A
Hybrid Approach of Computer Vision and
Sensor Network
Camera images
and Sensor
data
NN
Features of Soil images were extracted from the images
captured using the DSLR camera and moisture level of the
soils were measured using a sensor. All these data was used
in classification of soil types.
[151999
An artificial neural network for classifying
and predicting soil moisture and temperature
using Levenberg-Marquardt algorithm.
Sensor data
NN with
Levenberg-
Marquardt
Classified two soil classes grass and bare smooth soils based
on the remotely sensed data to measure the moisture of the
soil. Measured the moisture levels at different depths for the
same soil samples and classify the soil types.
[16]2017
Performance of SVM classifier for image
based soil classification.
Android and
SLR Images
SVM
Classified the soil types by extracting the features using LPF,
Gabor filter bank, and color momentum of the soil images
obtained from online mobile images
[17]2020
Soil texture classification using multi class
support vector machine
Texture features of the labelled soil samples were extracted
by using HSV histogram, color moment, Gabor wavelets,
and DWT. The features were classified using SVM
classifier. The soil samples were collected from paddy fields
in Guwahati.
[18]2020
Artificial intelligence system for supporting
soil classification.
CNN
Classified the soils sand, clay and gravel using SLR camera
images
Research Gaps
 Even though many promising methodologies have been developed and implemented,
research on classification of soil type using mobile/SLR camera images is in initial
stages.
 Classification of soil type using mobile/SLR camera images is a challenging task.
 There is no publicly available soil image dataset created by following the standards in
capturing the images of the soil type.
 Classifying the soil type using the texture features of the soil type is complicated but
might give accurate results as the RGB characteristics of the soil vary from region to
region due to climate and environmental conditions.
 Soil classification using images can reduce the computational time and cost in
analyzing the soil type.
 Soil classification will help in soil management in agriculture sector.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 17
Crop Leaf Disease Classification
 The crop management is an important task at every stage of the of the crop growth.
 The crop management involves
Yield Prediction
Crop quality
Leaf disease identification
Weed detection
Crop recognition
 To meet the necessity of the people, the production of the crop has to be increased.
 One of the most important feature in crop management is disease classification.
 Early detection of diseases in plants can reduce the pesticides and fertilizers usage in
protecting and increasing the yield of the crop.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 18
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 19
Fig 6. Percentage of primary crops production across the world. [19].
Most investigated crops[29]
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 20
Fig 7. Research article count on the primary crop [19]
 Potato (Solanum Tuberosum) has been chosen for plant leaf disease detection.
 Potato stands top in most consuming vegetable across the globe.
 India stands second largest producer nation, with 53.69 million metrics tons produced
during the fiscal year 2021.
 The average consumption of potato by a person is 49.4 pounds during the year 2019
[20].
 Potato leaves are affected due to fungal pathogens is
Early Blight (Alternaria)
Late Blight (Phytophthora)
 The annual potato crop loss due to late blight alone is 6.7 billion dollars across the
world.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 21
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 22
Early Blight Late Blight
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 23
Literature Review on Potato Leaf Disease
Classification
[Ref]
Year
Title Input Data Methods Focus
[21]2018
Using Support Vector Machines classification to
differentiate spectral signatures of potato plants
infected with Potato Virus Y
Hyperspectral
images
SVM
Classification of potato virus Y infected plants
using unmanned aerial vehicle to capture the
hyperspectral images with a spectral range
350-2500 nm in NIR and SWIR wavelengths.
The spatial samples were classified using SVM
[22]2021
An Empirical Study on Machine Learning Models
for Potato Leaf Disease Classification using RGB
Images
Central Potato
Research
Institute
(CPRI), India,
Plantvillage
Gabor filter,
Fine tuned
VGG16,SVM
Collected the real time images from the field
and segmented the region of interest and the
features are extracted using FCNN which are
used in ML for classification using SVM.
[23]2021
Automated abnormal potato plant detection
system using deep learning models and portable
video cameras
Video frame
images
CNN
Abnormal potato plants were identified during
the early growth and mid-growth of the plant
[24]2020
Detection of Potato Disease Using Image
Segmentation and Multi class Support Vector
Machine
Plantvillage
GLCM, Hu,
Color
Histogram, ML
Feature extraction from the diseased
segmented part and classifying the leaves
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 24
[Ref]
Year
Title Input Data Methods Focus
[25]2021
Ensemble Classification and Feature
Extraction Based Plant Leaf Disease
Recognition
Plantvillage
Dataset
GLCM+LBP+PCA
+LAW texture
mask+LDA+ML
Ensemble methods are created for extracting the features
using different sizes convolution mask and the features
are used in classifying the leaf disease images of Tomato,
Bell Pepper and Potato
[26]2021
Recognition of early blight and late
blight diseases on potato leaves based
on graph cut segmentation
Graph cut + LBP +
ML (SVM)
For Classifying the potato leaf diseases at different
stages, dataset has been segregated and segmented the
images using Graph cut model to detect the diseases. The
features of the images are extracted by using LBP and fed
to ML classifier algorithms.
[27]2021
Automated plant leaf disease detection
and classification using optimal
MobileNet based convolutional neural
networks
Extreme learning +
ML
Classified the tomato plant leaf diseases by optimizing
the hyperparameters of the model using EPO algorithm
[28]2022
Transfer Learning for Multi-Crop Leaf
Disease Image Classification using
Convolutional Neural Network VGG
VGG16
Multi crop leaf disease classification at early stages in
tomato and grape crop leaf. The diseased spot is
annotated by rectangular bounding box.
[29]2020
Identification of Disease in Potato
Leaves Using Convolutional Neural
Network (CNN) Algorithm
CNN
Classification of potato leaf diseases Early Blight and
Late Blight
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 25
Research Gap
 Research in identifying the diseased leaves in potato plants is still in early stages.
 Classification is carrying on the laboratory images for developing the models.
 Segmentation of infected leaf in identifying the diseases is a challenging task.
 Availability of real time images is low.
 There is no cost effective infrastructure in monitoring the crop disease leaves.
Problem Statement
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 26
1. Creating the standard soil image database.
2. Classifying the standard soil types sand and clay using machine learning algorithm
3. Multi-class soil classification using the soil image database created by considering
the imbalance data.
4. Classification of Phytophthora infestans and Alternaria solani diseases in Potato
plant leaves using deep learning networks.
5. Detection of diseases caused by fungi pathogens in Solanum Tuberosum leaves.
20-05-2023 Department of ECE, NIT Silchar 27
Objective 1
The standard soil image database has been created by capturing the images of the soil sample using
android/SLR camera.
Classifying the standard soil types sand and clay using machine learning algorithm
Objective 2
Multi-Feature Fusion for Soil Image Feature Extraction and Classification using Machine Learning
Multi-class soil classification using the soil image database created by considering the imbalance data
Objective 3
A Smart Soil Image Classification System using Lightweight Convolutional Neural Network
Classification of Phytophthora infestans and Alternaria solani diseases in Potato plant leaves using
deep learning network
Objective 4
POT-Net: Solanum Tuberosum (Potato) Leaves Diseases Classification using an Optimized Deep
Convolutional Neural Network
Detection of diseases caused by fungi pathogens in Solanum Tuberosum leaves
Objective 5
Early Blight and Late Blight disease detection using deep learning models
Creating the standard soil image database
Problem Statement 1
Objective
 The standard soil image database has been created by capturing the images of the soil sample
using android/SLR camera.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 28
Acquiring Soil Samples
 The soil samples are collected from different regions of Andhra Pradesh and Silchar, Assam.
 A total of 96 soil samples are gathered from agricultural fields in these locations.
 The regions include the Godavari river coastal zone, rich in alluvial and sandy soils, drought-
prone areas with sandy and rocky soils, and hill stations suitable for plantation.
 The soil samples were obtained at a depth of 5 to 10 cm below the field surface.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 29
(a) (b) (c)
Fig. 8. Represents the areas where soil samples are collected. (a) Rajahmundry (b) Madanapalle (c) Digged area
Soil analysis in the laboratory
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 30
 Each soil sample is subjected to laboratory examination to determine the texture of the
collected soil samples and label the soil type by identifying the fraction of sand, silt, and clay.
 Sieve Analysis
 Hydrometer test
• In sieve analysis, 500gms of sand type soils are taken and
sieved through the different sieve pans of sieve size up to
75μm to find the fraction of sand. If the fraction of silt and
clay that passed through the 75μm sieve pan is greater than 10
percent, we go for a hydrometer test to find the silt and clay
fraction.
Fig.9. Sieve shaker for Sieve Analysis
 For hydrometer test, 33gms of sodium hexametaphosphate and 7gms of sodium bicarbonate
in one liter of water, a solution is made and left for 24 hours.
 Now, 100ml of the solution is mixed with 800ml of water for the hydrometer test.
 In the 900ml prepared solution, 50gm of the tested soil sample is added, and readings are
taken at regular intervals over the following 24 hours.
 The height of the hydrometer from the surface of the solution is recorded at regular intervals.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 31
• The texture of the soil sample is determined
by plotting the percentage of sand, silt and
clay on the United States Department of
Agriculture (USDA) texture triangle.
(a) (b)
Fig. 10. (a) Hydrometer and (b) Hydrometer test
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 32
(a) (b)
Fig 11. a) USDA texture triangle for classification. b) USDA texture triangle with the some soil samples plotted.
Experimental Setup
 The images of the acquired soil samples are captured by using a Samsung smartphone
camera.
 The organic components present in the soil are removed, and the moisture content is removed
by drying the soil in a hot air oven at 1000C for 24 hours.
 A chamber with dimensions 45x20x10 cm was constructed with Styrofoam to capture the
images.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 33
• As the box walls are white, allowing the
light that enters the box through the
window will provide the illumination
conditions enough to capture the soil
images without affecting the actual color
of the soil.
(a) (b)
Fig 12. Image acquisition setup. a) Isometric view of a chamber (b) Styrofoam
box to capture images.
 The soil images are captured using a smartphone camera with 48 megapixels.
 A total of 392 soil images were captured from the collected samples.
 The dataset is made available in the IEEE data portal.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 34
(a) (b) (c)
Fig 13. Represents the clay, loamy sand, and sandy loam soil images captured using a smartphone camera.
Summary
 This work presents the creation soil image database.
 The soil image database created by following standards so that the soil
samples can be labelled accurately.
 The database may help the researchers in identifying the soil types by
image analysis.
 This can enable in advancement of agriculture towards automation.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 35
Objective
 Multi-Feature Fusion for Soil Image Feature Extraction and Classification using
Machine Learning
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 36
Problem Statement 2
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 37
• Contributions
• Initially 70 soil samples were collected from different regions
of Andhra Pradesh
• The feature fusion technique has been proposed to extract the
soil texture features.
• The GSVM machine learning classifier has been trained to
classify the soil types by varying the kernel parameter.
(a) (b) (c)
Fig. 14. ROI extracted images (a) Coarse sand (b) Fine sand (c) Clay soil
images.
Fig.15. Work Flow of the model
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 38
𝐺𝐿𝐶𝑀 =
𝑝 1,1 𝑝 1,2 ⋯ 𝑝 1, Mg
𝑝 2,1 𝑝 2,2 ⋯ 𝑝 2, Mg
⋮ ⋮ ⋱ ⋮
𝑝 Mg, 1 𝑝 Mg, 2 ⋯ 𝑝 Mg, Mg
. . 1
• Calculate the features: contrast, correlation, energy,
homogeneity, and entropy using GLCM [30].
• Calculate the Tamura features [31].
• Gabor filter bank is created by using two scales, six
orientations, and four frequency values [32].
• The ROI images are quantized with hue, saturation, and
value with 6 × 3 × 5 equal bins. The output is a vector
of size 1 × 90.
• The feature fusion vector size for each image is 1 ×
202.
(a) (b)
(c) (d)
Fig 16. a) Confusion matrix of the proposed model. (b) True positive and False-negative rate
of the model. (c) ROC curve of clay and (d) ROC curve of sand.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 39
(a) (b) (c)
(d) (e) (f)
Fig.17. Confusion matrix of GSVM using (a) GLCM (b) Tamura (c) T+GLCM
(d) Gabor (e) Gabor+GLCM (f) Feature fusion model
State-of-
the-art GLCM Tamura
T+
GLCM Gabor
Gabor +
GLCM
Feature
Fusion
GSVM [33] 85 82.5 80 85 95 97.5
Table 1. Ablation experimental analysis
State-of-
the-art GLCM Tamura
T+
GLCM Gabor
Gabor+
GLCM
Feature
Fusion
LR [34] 65 60 90 72.5 62.5 57.5
NB [36] 55 55 57.5 57.5 62.5 70
WKNN [35] 90 62.5 77.5 77.5 90 90
BNN [37] 95 65 72.5 72.5 70 92.5
SVM 42.5 60 52.5 67.5 62.5 87.5
GSVM 85 82.5 80 85 95 97.5
Table 2. Comparative analysis of the state-of-the-art in texture
features and ML classification algorithms
Algorithm Precision Recall Specificity F1-score
Error
rate
Accuracy
(%)
LR 0.55 0.578 0.571 0.564 0.85 57.5
NB 0.8 0.67 0.75 0.727 0.6 70
GSVM 0.95 1 0.952 0.974 0.05 97.5
WKNN 0.9 0.9 0.9 0.9 0.1 90
SVM 0.9 0.857 0.895 0.878 0.125 87.5
B-NN 1 0.869 1 0.929 0.15 92.5
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 40
Fig. 18. Confusion Matrix of (a) LR (b) NB (c) SVM (d) WKNN
(e) B-NN (f) GSVM.
(a) (b) (c)
(d) (e) (f)
Table 3. Performance analysis in classifying the soils with state-of-the-art
ML algorithms.
Fig. 19. Comparison of the proposed model with state-of-the-art using 10-fold cross-
validation
Summary
 This work presents multi-feature fusion model for the classification of standard soils
sand and clay using Gaussian SVM.
 The feature fusion model is compared with state-of-the-art machine learning
classifiers.
 This can enable the researchers in advancement and upgrade of techniques for the
classification various soil types.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 41
Objective
Smart Soil Image Classification System using Lightweight Convolutional Neural
Network
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 42
Problem Statement 3
Contributions
 Rather than classifying the gravel and aggregate soils, classified the soil types like
sand, clay, loam, sandy loam and loamy sand related to agriculture soils.
 HSV, RGB extraction, and adaptive histogram techniques are applied to the database
to highlight the texture features of the soil sample images.
 Developing a novel Lightweight network aimed to classify the soil images with less
number of layers, learnable parameters, epochs, size of the network, and better
performance.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 43
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 44
Fig. 20. The workflow of the proposed model in the classification of soil.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 45
(a) (b) (c)
Fig. 21. a, b, and c represent the clay, loamy sand and sand images and the corresponding ROI images.
(a) (b) (c) (d) (e)
Fig. 22. Sample Pre-processed images. (a) Adaptive histogram [38]. (b-d) RGB channels respectively. [39] (e) V extraction
from HSV.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 46
Fig. 23. Proposed Light-SoilNet network architecture
Class TP TN FP FN Precision Recall F1 score
Clay 146 430 0 2 1.000 0.986 0.993
Loam 90 478 10 0 0.900 1.000 0.947
Sandy
Loam
100 465 0 13 1.000 0.885 0.939
Loamy
Sand
121 454 5 2 0.960 0.984 0.972
Sand 104 476 0 2 1.000 0.981 0.990
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 47
Adaptive
Histogram
V
value
RGB
Channels Accuracy
F1 Score
Clay Loam Loamysand Sand Sandyloam
-   81.4 0.83 0.92 0.95 0.64 0.82
 -  78.4 0.79 0.50 0.92 0.86 0.55
  - 92.8 0.94 0.75 0.88 0.95 1.00
   97.2 0.99 0.94 0.97 0.99 0.94
Table 5. Ablation experiment results for preprocessed techniques.
Fig. 24. Confusion matrix and ROC of the Light-SoilNet network.
Table 4. Performance metrics of the proposed model for each class.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 48
Architecture Accuracy (%) Size (MB)
Proposed model 97.20 9.62
MobileNet-v2 [40] 95.45 21.5
MobileNet-v3 94.5 21.1
EfficientNet-B0 [41] 87.54 28.3
ShuffleNet [42] 95.67 14.5
Network
Clay Loam Sand Sandy loam Loamy sand
P R P R P R P R P R
MobileNet-v2 0.918 0.944 1 0.980 0.885 0.920 1 1 1 0.940
ShuffleNet 0.973 0.986 1 1 0.906 1.000 1 0.901 0.913 0.906
EfficientNet-
B0
0.849 0.756 0.500 1 1 0.779 1 1 1 0.984
Light-SoilNet 1 0.986 0.900 1 1 0.981 1 0.885 0.960 0.984
Table 6. Classification accuracy of Light-SoilNet and pre-trained lightweight networks.
(a) (b)
(c) (d)
Fig. 25. Confusion matrix. (a) EfficientNet-B0 (b) MobileNet-v2 (c) ShuffleNet (d) MobileNet-v3.
Table 7. Performance analysis for five soil classes with pre-trained lightweight networks.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 49
Architecture
Accuracy
(%)
Parameters
(Millions)
FLOPs
(G)
Inference
Time (sec)
Training
Time (min)
Size
(MB)
Proposed model 97.20 2.7 3.74 11 16 9.62
Inception-v3 [44] 94.46 24 5.72 6.5 6.5 94.1
Vgg-19 [43] 85.12 144 15.47 7 7.1 519
ResNet-50 [43] 94.46 23 3.87 4.5 6 100
AlexNet [45] 91.00 61 7.27 4.1 7.4 226
Network
Clay Loam Sand Sandy loam Loamy sand
P R P R P R P R P R
AlexNet 0.822 0.923 0.900 1 0.981 0.754 1.000 1 0.889 0.982
Inception 0.918 1 1 1 1 0.779 0.800 1 1 0.984
Vgg-19 0.959 0.833 1 1 0.528 0.651 0.700 1 1 0.685
ResNet-50 1 0.820 1 1 0.962 1 1 1 0.778 1.000
Light-
SoilNet
1 0.986 0.900 1 1 0.981 1 0.885 0.960 0.984
Table 8. Classification accuracy with comparative results.
(a) (b)
(c) (d)
Fig. 26. Confusion matrix. (a)VGG-19 (b)AlexNet (c) ResNet-50 (d) Inception-v3.
Table 9. Performance analysis for the five soil classes with pre-trained DL networks.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 50
Fig. 27. Comparative analysis of F1-score for the soil images.
Fig. 28. Comparison of Light-SoilNet with the pre-trained DL and
lightweight networks using accuracy as cross-validation.
Model
IRSID
Dataset
Accuracy
(%)
Online
Dataset
Accuracy
(%)
(Azizi, 2020)ResNet-50 94.46 68.1
(Azizi, 2020)Vgg-16 92.8 57.4
(Azizi, 2020)Inception-v3 94.46 68.09
(Inazumi, Ph, 2020) 92.72 70.2
Proposed (Light-SoilNet) 97.2 65.7
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 51
Table 10. Comparison of the Light-SoilNet and IRSID
dataset with existing models and online dataset.
Summary
• This paper presents a new lightweight CNN-based network
architecture for classifying the five different types of soil images.
• The performance of the model compared with pre-trained
lightweight and DL pre-trained networks.
• The proposed Light-SoilNet architecture outperforms the state-of-
the-art models in classifying the soil images in terms of accuracy,
reducing the learnable parameters, FLOPs, and memory of the
architecture.
• This can enable the researchers in advancement and upgrade of
techniques for the classification various soil types using standard
images.
Objective
 POT-Net: Solanum Tuberosum (Potato) Leaves Diseases Classification using an
Optimized Deep Convolutional Neural Network
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 52
Problem Statement 3
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 53
Fig 29. Schematic diagram of the proposed POT-Net model
Size/Operation Activation Filter Depth Stride Number of
Parameters
Input (224×224×3) - - - - 0
CL_1 224×224×3 7×7 128 1 18944
BNL_1+ReLU_1 224×224×128 - - - 256
Max Pool_1 112×112×128 2×2 - 2 0
CL_2 112×112×128 5×5 128 1 409728
BNL_2+ReLU_2 112×112×128 - - - 256
Max Pool_2 56×56×128 2×2 - 2 0
CL_3 56×56×64 5×5 64 1 204864
BNL_3+ReLU_3 56×56×64 - - - 128
Max Pool_3 28×28×64 2×2 - 2 0
CL_4 28×28×64 3×3 64 1 36928
BNL_4+ReLU_4 28×28×64 - - - 128
Max Pool_4 14×14×64 2×2 - 2 0
CL_5 14×14×32 3×3 32 1 18464
BNL_5+ReLU_5 14×14×32 - - - 64
Max Pool_5 7×7×32 2×2 - 2 0
CL_6 7×7×16 3×3 16 1 4624
BNL_6+ReLU_6 7×7×16 - - - 32
Max Pool_6 3×3×16 2×2 - 2 0
CL_7 3×3×16 3×3 16 1 2320
BNL_7+ReLU_7 3×3×16 - - - 32
Max Pool_7 1×1×16 2×2 - 2 0
CL_8 1×1×8 3×3 8 1 1160
BNL_8+ReLU_8 1×1×8 - - - 16
FC 1×1×3 - - - 27
Softmax 1×1×3 - - - 0
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 54
Contribution
• A novel Optimized deep learning network to classify Potato leaf
diseases using CNN architecture (POT-Net) has been developed to
identify the EB and LB diseased leaves of the potato plant from
healthy leaves with less run-time.
• The proposed model has an optimum number of parameters
compared with the state-of-the-art DL techniques, which reduces
the cost of the model.
• The hyperparameters of the model are optimized using a
metaheuristic algorithm called Whale Optimization Algorithm by
improving the efficiency of the model in identifying the diseased
leaves with an accuracy of 99.12%.
• The proposed model performance has been compared with the
state-of-the-art models, pre-trained DL networks and various meta-
heuristic optimization algorithms to prove the efficiency of the
model.
Table 11. Proposed CNN Architecture with parameters and filter values at each layer.
 Encircling the prey:
𝐷 = 𝐶. 𝑋𝑟 𝑡 − 𝑋(𝑡) … (7)
𝑋 𝑡 + 1 = 𝑋𝑟 𝑡 − 𝐴 ∙ 𝐷 … (8)
 If 𝐴 ≥ 1, then the search agent is far from the optimal solution, and a new search
agent is chosen.
 . If 𝐴 < 1 best solution is achieved by the search agent.
 Exploitation Phase:
𝑋 𝑡 + 1 =
𝑋𝑖 𝑡 − 𝐴 ∙ 𝐷, 𝑖𝑓 𝑝 < 0.5 (𝑆ℎ𝑟𝑖𝑛𝑘𝑖𝑛𝑔 𝑒𝑛𝑐𝑖𝑟𝑙𝑒)
𝐷′
∙ 𝑒𝑠𝑙
∙ cos 2𝜋𝑙 + 𝑋𝑖 𝑡 , 𝑖𝑓 𝑝 ≥ 0.5 (𝑆𝑝𝑖𝑟𝑎𝑙 𝑢𝑝𝑑𝑎𝑡𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛)
… (12)
𝐷
′
= 𝑋𝑖 𝑡 − 𝑋(𝑡)
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 55
Whale Optimization Algorithm (WOA) [46]
Fig. 30. Creating spiral shape bubbles by shrinking
the circles to catch the prey.
Model Momentum Epochs L2regularization Learning rate
CNN optimized with WOA 0.7070 6.3717 0.0002 0.0892
CNN without optimization 0.01 8 0.0001 0.01
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 56
Class TP TN FP FN Precision Recall F1 score
Early Blight 1056 1772 14 4 0.996 0.987 0.991
Healthy 730 2109 7 0 1 0.99 0.995
Late Blight 1035 1786 4 21 0.980 0.996 0.988
(a) (b) (c)
Fig 31. Sample images. (a) Healthy (b) Early Blight (c) Late Blight [47]
Category No. of images
Training set Validation set
Early Blight 2472 1060
Healthy 1702 730
Late Blight 2465 1056
Total 6639 2846
Table 12. Dataset Description for training and validation.
(a) (b)
Results
Fig 32. POT-Net model (a) Confusion matrix (b) ROC
Table 14. POT-Net model performance parameters and metrics.
Table 13. Training options of the CNN with and without optimization
Model Accuracy (%)
Learnable Parameters
(millions)
Computation time
per image (sec)
Sholihati et al. [48] 91.31 22 0.34
Deepa et al. [49] 88 - 0.6
Network1 93.64 0.3 0.08
Network2 97.26 1.56 0.15
POT-Net 99.12 0.7 0.12
Table 15. Performance analysis with computational time and learnable parameters.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 57
Architecture
Accuracy
(%)
Parameters
(Millions)
AlexNet [45] 97.79 61
DenseNet [33] 96.21 7.2
ResNet-50 [43] 91.64 23
MobileNet-v2 [40] 90.55 3.47
ShuffleNet [42] 98.63 2.29
EfficientNet-B0 [41] 98.77 26
VGG-19 [43] 98.7 144
Non-Opt CNN 97 0.7
POT-Net 99.12 0.7
Network
Early Blight Healthy Late Blight
Precision Recall F1-score Precision Recall F1-score Precision Recall
F1-
score
AlexNet 0.988 0.982 0.985 0.953 0.998 0.975 0.985 0.96 0.972
DenseNet 0.969 0.986 0.977 0.957 0.955 0.956 0.959 0.941 0.949
ResNet-50 0.968 0.934 0.951 0.898 0.915 0.906 0.875 0.898 0.886
MobileNet 0.875 0.971 0.921 0.993 0.839 0.909 0.876 0.899 0.887
ShuffleNet 0.982 0.992 0.986 0.986 0.995 0.99 0.991 0.974 0.982
EfficientNet-B0 0.982 0.986 0.984 0.997 0.998 0.997 0.986 0.981 0.983
VGG-19 0.981 0.992 0.986 0.991 0.991 0.991 0.989 0.978 0.983
Non-Opt CNN 0.974 0.995 0.984 1 0.921 0.958 0.947 0.985 0.965
POT-Net 0.996 0.987 0.991 1 0.99 0.995 0.98 0.996 0.988
Performance analysis with pre-trained DL and non-optimized CNN network
Table 16. Classification accuracy of POT-Net and pre-trained
networks. Table 17. Performance metrics of the pre-trained DL networks.
Optimizer Accuracy (%)
GWO [39] 98.35
SMA [40] 98.91
PSO [41] 97.29
POT-Net 99.12
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 58
Performance analysis with MHO algorithms
Optimizer
Early Blight Healthy Late Blight
Precision Recall F1-score Precision Recall F1-score Precision Recall F1-score
GWO 0.983 0.988 0.985 0.981 0.994 0.988 0.986 0.972 0.979
SMA 0.987 0.989 0.988 1.000 0.992 0.996 0.984 0.988 0.986
PSO 0.973 0.982 0.977 0.996 0.953 0.974 0.957 0.979 0.968
POT-Net 0.996 0.987 0.991 1 0.99 0.995 0.98 0.996 0.988
Table 18. Optimizer algorithms performance with proposed CNN architecture.
Table 19. Performance metrics of the proposed CNN with the optimizer algorithms.
(a) (b)
(c)
Fig. 33.. Confusion matrix. (a) GWO (b) SMA (c) PSO.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 59
(a) (b)
Fig. 34. Testing the POT-Net model with real-time images. (a) Confusion matrix (b) ROC
The proposed model has been validated with real-time
images containing 395 images. The model has identified
the EB, LB and healthy leaves images with an accuracy of
81%.
Summary
 This paper uses image phenotyping to classify the potato plant leaves when infected with diseases.
 An image-based framework for identifying the fungal pathogen disease leaves using a CNN
architecture has been presented in this article.
 The proposed POT-Net model performance is evaluated with accuracy, precision, recall, and F1-score
and compared with the pre-trained DL networks and state-of-the-art optimization algorithms.
 The model has achieved high performance in classifying the accuracy is 99.12%, and the time to
process an image to classify the diseased leaves is 0.12 sec.
 In future, by exploring different CNN architectures, the performance will be extended in identifying the
leaf diseases of potato and other plant leaf diseases using laboratory and field images.
 the proposed methodology might be viewed as an effective method for the early detection of diseased
and healthy leaves in potato plants in minimizing the significant agriculture loss, which helps in
automation and continuous monitoring of the fields.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 60
 Digital transformation in Agriculture can meet the food demand of the global population with less
workforce.
 Soil type classification can increase the crop yield by choosing the site specific crops and automation
of the agriculture sector.
 Standard soil image database has been created to classify the soil types using various machine learning
and deep learning approaches by increasing the classification accuracy and minimizing the complexity
of the model.
 To increase the yield of the crop, detecting the plant fungal diseases at early stages was achieved by
using deep learning approaches.
 The performance of the proposed models are compared with other state-of-the-art techniques.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 61
Conclusion
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 62
Publications
Conferences
1. D. N. K. Pandiri, R. Murugan and T. Goel, "ODNet: Optimized Deep Convolutional Neural Network for
Classification of Solanum Tuberosum Leaves Diseases," 2022 IEEE Region 10 Symposium (TENSYMP),
Mumbai, India, 2022, pp. 1-6, doi: 10.1109/TENSYMP54529.2022.9864335
Database
1. D N Kiran Pandiri, R Murugan, Tripti Goel, March 18, 2021, "Indian Regions Soil Image Database
(IRSID): A dataset for classification of Indian soils", IEEE Dataport, doi:
https://dx.doi.org/10.21227/2zz3-f173
Journals
1. D. N. Kiran Pandiri, R. Murugan, Tripti Goel, Nishant Sharma, Aditya Kumar Singh, Soumya Sen &
Tonmoy Baruah (2023) POT-Net: solanum tuberosum (Potato) leaves diseases classification using an
optimized deep convolutional neural network, The Imaging Science
Journal, DOI: 10.1080/13682199.2023.2169988 (SCIE, I.F – 0.871)
Works communicated
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 63
Journals
1. D N Kiran Pandiri, R. Murugan, T. Goel “Multi-Feature Fusion for Soil Image Feature Extraction and
Classification using Machine Learning,” Signal Image and Video Processing (under Review).
2. D N Kiran Pandiri, R. Murugan, T. Goel “Smart Soil Image Classification System using Lightweight
Convolutional Neural Network,” Expert Systems with Applications (Under Review).
3. D N Kiran Pandiri, R.Murugan, T. Goel “Classification of Soil using Machine Learning and Deep
Learning Techniques–A Review,” IEEE Access (Under Review).
References
1. https://www.downtoearth.org.in/factsheet/how-livestock-farming-affects-the-environment-64218
2. https://www.fao.org/3/cb1329en/online/cb1329en.html#chapter-1
3. FAO. 2020. World Food and Agriculture - Statistical Yearbook 2020. Rome.
https://doi.org/10.4060/cb1329en
4. https://www.sciencedirect.com/topics/earth-and-planetary-sciences/soil-management
5. Rajapakse, R., 2015. Geotechnical Engineering Calculations and Rules of Thumb: Second Edition, Geotechnical Engineering
Calculations and Rules of Thumb: Second Edition. Butterworth-Heinemann.
6. Meigh, A.C., 2013. Cone Penetration Testing: Methods and Interpretation. Elsevier.
7. Bouyoucos, G.J., 1962. Hydrometer Method Improved for Making Particle Size Analyses of Soils 1 . Agron. J. 54, 464–465.
https://doi.org/10.2134/agronj1962.00021962005400050028x
8. Lu, N., Ristow, G.H., Likos, W.J., 2000. The Accuracy of Hydrometer Analysis for Fine-Grained Clay Particles. Geotech. Test. J. 23,
487–495. https://doi.org/10.1520/gtj11069j
9. Vibhute, A.D., Kale, K.V., Dhumal, R.K. and Mehrotra, S.C., 2015, December. Soil type classification and mapping using hyperspectral
remote sensing data. In 2015 International Conference on Man and Machine Interfacing (MAMI) (pp. 1-4). IEEE.Vibhute, A.D., Kale,
K.V., Dhumal, R.K. and Mehrotra, S.C., 2015, December. Soil type classification and mapping using hyperspectral remote sensing data.
In 2015 International Conference on Man and Machine Interfacing (MAMI) (pp. 1-4). IEEE.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 64
10. Yu, Y., Xu, T., Shen, Z., Zhang, Y., Wang, X., 2019. Compressive spectral imaging system for soil classification with three-
dimensional convolutional neural network. Opt. Express 27, 23029. https://doi.org/10.1364/oe.27.023029
11. Mengistu, A.D. and Alemayehu, D.M., 2018. Soil Characterization and Classification: A Hybrid Approach of Computer Vision
and Sensor Network. International Journal of Electrical & Computer Engineering (2088-8708), 8(2).
12. Azizi, A., Gilandeh, Y.A., Mesri-Gundoshmian, T., Saleh-Bigdeli, A.A. and Moghaddam, H.A., 2020. Classification of soil
aggregates: A novel approach based on deep learning. Soil and Tillage Research, 199, p.104586.
13. Chung, S.O., Cho, K.H., Cho, J.W., Jung, K.Y., Yamakawa, T., 2012. Soil texture classification algorithm using RGB
characteristics of soil images. J. Fac. Agric. Kyushu Univ. 57, 393–397. https://doi.org/10.3182/20101206-3-jp-3009.00005
14. Mengistu, A.D. and Alemayehu, D.M., 2018. Soil Characterization and Classification: A Hybrid Approach of Computer Vision
and Sensor Network. International Journal of Electrical & Computer Engineering (2088-8708), 8(2).
15. Atluri, V., Hung, C.C. and Coleman, T.L., 1999, March. An artificial neural network for classifying and predicting soil moisture
and temperature using Levenberg-Marquardt algorithm. In Proceedings IEEE Southeastcon'99. Technology on the Brink of
2000 (Cat. No. 99CH36300) (pp. 10-13). IEEE.
16. Srunitha, K., Padmavathi, S., 2017. Performance of SVM classifier for image based soil classification. Int. Conf. Signal
Process. Commun. Power Embed. Syst. SCOPES 2016 - Proc. 411–415. https://doi.org/10.1109/SCOPES.2016.7955863
17. Barman, U. and Choudhury, R.D., 2020. Soil texture classification using multi class support vector machine. Information
processing in agriculture, 7(2), pp.318-332.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 65
18. Inazumi, S., Ph, D., Intui, S., Eng, M., Jotisankasa, A., Ph, D., Chaiprakaikeow, S., Ph, D., Kojima, K., 2020. Artificial
intelligence system for supporting soil classification. Results Eng. 8. https://doi.org/10.1016/j.rineng.2020.100188
19. Benos, L., Tagarakis, A.C., Dolias, G., Berruto, R., Kateris, D. and Bochtis, D., 2021. Machine learning in agriculture: A
comprehensive updated review. Sensors, 21(11), p.3758.
20. https://www.ers.usda.gov/data-products/chart-gallery/gallery/chart-detail/?chartId=58340
21. Griffel, L.M., Delparte, D. and Edwards, J., 2018. Using Support Vector Machines classification to differentiate spectral
signatures of potato plants infected with Potato Virus Y. Computers and electronics in agriculture, 153, pp.318-324.
22. Ghosh, S., Rameshan, R. and Dinesh, D.A., 2021. An Empirical Study on Machine Learning Models for Potato Leaf Disease
Classification using RGB Images. In ICPRAM (pp. 515-522).
23. Oishi, Y., Habaragamuwa, H., Zhang, Y., Sugiura, R., Asano, K., Akai, K., Shibata, H. and Fujimoto, T., 2021. Automated
abnormal potato plant detection system using deep learning models and portable video cameras. International Journal of
Applied Earth Observation and Geoinformation, 104, p.102509.
24. Islam, M., Dinh, A., Wahid, K. and Bhowmik, P., 2017, April. Detection of potato diseases using image segmentation and
multiclass support vector machine. In 2017 IEEE 30th canadian conference on electrical and computer engineering
(CCECE) (pp. 1-4). IEEE.
25. N. Kaur and V. Devendran, "Ensemble Classification and Feature Extraction Based Plant Leaf Disease Recognition," 2021 9th
International Conference on Reliability, Infocom Technologies and Optimization (Trends and Future Directions) (ICRITO),
2021, pp. 1-4, doi: 10.1109/ICRITO51393.2021.9596456.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 66
26. Hou, C., Zhuang, J., Tang, Y., He, Y., Miao, A., Huang, H. and Luo, S., 2021. Recognition of early blight and late blight
diseases on potato leaves based on graph cut segmentation. Journal of Agriculture and Food Research, 5, p.100154.
27. Ashwinkumar, S., Rajagopal, S., Manimaran, V. and Jegajothi, B., 2021. Automated plant leaf disease detection and
classification using optimal MobileNet based convolutional neural networks. Materials Today: Proceedings.
28. Paymode, A.S. and Malode, V.B., 2022. Transfer learning for multi-crop leaf disease image classification using convolutional
neural networks VGG. Artificial Intelligence in Agriculture.
29. Rozaqi, A.J. and Sunyoto, A., 2020, November. Identification of Disease in Potato Leaves Using Convolutional Neural Network
(CNN) Algorithm. In 2020 3rd International Conference on Information and Communications Technology (ICOIACT) (pp. 72-
76). IEEE.
30. Suresh A, Shunmuganathan KL (2012) Image texture classification using gray level co-occurrence matrix based statistical
features. Eur J Sci Res 75:591–597
31. Tamura H, Mori S, Yamawaki T (1978) Textural Features Corresponding to Visual Perception. IEEE Trans Syst Man Cybern
8:460–473. https://doi.org/10.1109/TSMC.1978.4309999
32. Zhao M, Qiu W, Wen T, et al (2021) Feature extraction based on Gabor filter and Support Vector Machine classifier in defect
analysis of Thermoelectric Cooler Component. Comput Electr Eng 92:107188.
https://doi.org/10.1016/j.compeleceng.2021.107188
33. Fischetti M (2016) Fast training of Support Vector Machines with Gaussian kernel. Discret Optim 22:183–194.
https://doi.org/10.1016/j.disopt.2015.03.002
34. Fan Y, Bai J, Lei X, et al (2020) Privacy preserving based logistic regression on big data. J Netw Comput Appl 171:102769.
https://doi.org/10.1016/j.jnca.2020.102769
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 67
35. Dubey H, Pudi V (2013) Class based weighted K-Nearest neighbor over imbalance dataset. Lect Notes Comput Sci (including
Subser Lect Notes Artif Intell Lect Notes Bioinformatics) 7819 LNAI:305–316. https://doi.org/10.1007/978-3-642-37456-2_26
36. Balaji VR, Suganthi ST, Rajadevi R, et al (2020) Skin disease detection and segmentation using dynamic graph cut algorithm
and classification through Naive Bayes classifier. Meas J Int Meas Confed 163:107922.
https://doi.org/10.1016/j.measurement.2020.107922
37. Ackora-prah J, Email T (2014) A Bilayer Feed – Forward Artificial Neural Network for Exchange Rate Prediction. Aust J Appl
Math 9478:1–5
38. Singh, K., Vishwakarma, D. K., Walia, G. S., & Kapoor, R. (2016). Contrast enhancement via texture region based histogram
equalization. Journal of Modern Optics, 63(15), 1444–1450. https://doi.org/10.1080/09500340.2016.1154194
39. Kumar, E. B., & Thiagarasu, V. (2018). Color channel extraction in RGB images for segmentation. Proceedings of the 2nd
International Conference on Communication and Electronics Systems, ICCES 2017, 2018-Janua(Icces), 234–239.
https://doi.org/10.1109/CESYS.2017.8321272
40. Michele, A., Colin, V., & Santika, D. D. (2019). Mobilenet convolutional neural networks and support vector machines for
palmprint recognition. Procedia Computer Science, 157, 110–117. https://doi.org/10.1016/j.procs.2019.08.147
41. Synced. (2017). ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices. Medium.
https://medium.com/syncedreview/shufflenet-an-extremely-efficient-convolutional-neural-network-for-mobile-devices-
72c6f5b01651
42. Chowdhury, N. K., Kabir, M. A., Rahman, Md. M., & Rezoana, N. (2021). ECOVNet: a highly effective ensemble based deep
learning model for detecting COVID-19. PeerJ Computer Science, 7, 1–25. https://doi.org/10.7717/PEERJ-CS.551
43. Victor Ikechukwu, A., S, M., R, D., & RC, S. (2021). ResNet-50 vs VGG-19 vs Training from Scratch: A comparative analysis
of the segmentation and classification of Pneumonia from chest x-ray images. Global Transitions Proceedings.
https://doi.org/10.1016/j.gltp.2021.08.027
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 68
44. Islam, M. S., Foysal, F. A., Neehal, N., Karim, E., & Hossain, S. A. (2018). IncePTB: A CNN based classification approach
for recognizing traditional Bengali games. Procedia Computer Science, 143, 595–602.
https://doi.org/10.1016/j.procs.2018.10.436
45. Alex Krizhevsky, Ilya Sutskever, & Geoffrey E. Hinton. (2012). ImageNet Classification with Deep Convolutional Neural
Networks. Advances in Neural Information Processing Systems, 25, 1097–1105.
46. S. Mirjalili, A. Lewis, The whale optimization algorithm, Adv. Eng. Softw. 95 (2016) 51–67.
47. Alex Lavaee, PlantifyDr Dataset | Kaggle, (n.d.). https://www.kaggle.com/lavaman151/plantifydr-dataset.
48. Sholihati RA, Sulistijono IA, Risnumawan A, et al. Potato Leaf Disease Classification Using Deep Learning Approach, In:
2020 Int. Electron. Symp., IEEE, 2020:pp. 392–397.
49. NR Deepa, N. Nagarajan, Kuan noise filter with Hough transformation based reweighted linear program boost classification
for plant leaf disease detection, J. Ambient Intell. Humaniz. Comput. 12 (2021) 5979–5992.
50. M. Canayaz, MH-COVIDNet: Diagnosis of COVID-19 using deep neural networks and meta-heuristic-based feature selection
on X-ray images, Biomed. Signal Process. Control. 64 (2021) 102257.
51. S. Li, H. Chen, M. Wang, A.A. Heidari, S. Mirjalili, Slime mould algorithm: A new method for stochastic optimization, Futur.
Gener. Comput. Syst. 111 (2020) 300–323.
52. A. Herliana, T. Arifin, S. Susanti, AB Hikmah, Feature Selection of Diabetic Retinopathy Disease Using Particle Swarm
Optimization and Neural Network, 2018 6th Int. Conf. Cyber IT Serv. Manag. CITSM 2018. (2019) 2016–2019.
https://doi.org/10.1109/CITSM.2018.8674295.
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 69
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 70
1st
Semester
2nd
Semester
3rd
Semester
4th
Semester
5th
Semester
6th
Semester
7th
Semester
8th
Semester
Coursework
completed
Comprehensive
Appeared and Literature
Survey started
Literature survey, problem
formulation
Literature survey,
Implementation of problem
statement 1
Implementation of
problem statements 1,
2 using MATLAB
Implementation of
problem statements 4
using MATLAB
Implementation of
problem statement 3
using MATLAB
Synopsis and Thesis
submission
9th
Semester
Implementation of
problem statement 5 using
MATLAB/Python
20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 71
Thank
You

More Related Content

Similar to Enhancement PPT.pptx

Applications of Remote Sensing
Applications of Remote SensingApplications of Remote Sensing
Applications of Remote SensingAbhiram Kanigolla
 
IRJET - Survey on Soil Classification using Different Techniques
IRJET - Survey on Soil Classification using Different TechniquesIRJET - Survey on Soil Classification using Different Techniques
IRJET - Survey on Soil Classification using Different TechniquesIRJET Journal
 
Land use Land Cover Highlight for Jibia Local Government, Nigeria
Land use Land Cover Highlight for Jibia Local Government, NigeriaLand use Land Cover Highlight for Jibia Local Government, Nigeria
Land use Land Cover Highlight for Jibia Local Government, Nigeriaijtsrd
 
International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)IJERD Editor
 
Precision agriculture
Precision agriculturePrecision agriculture
Precision agricultureSuryaBv1
 
IRJET - Normalized Difference Vegetation Index (NDVI) based Land Cover Cl...
IRJET -  	  Normalized Difference Vegetation Index (NDVI) based Land Cover Cl...IRJET -  	  Normalized Difference Vegetation Index (NDVI) based Land Cover Cl...
IRJET - Normalized Difference Vegetation Index (NDVI) based Land Cover Cl...IRJET Journal
 
An automated low cost IoT based Fertilizer Intimation System forsmart agricul...
An automated low cost IoT based Fertilizer Intimation System forsmart agricul...An automated low cost IoT based Fertilizer Intimation System forsmart agricul...
An automated low cost IoT based Fertilizer Intimation System forsmart agricul...sakru naik
 
Application of Remote Sensing GIS in Agriculture.pdf
Application of Remote Sensing   GIS in Agriculture.pdfApplication of Remote Sensing   GIS in Agriculture.pdf
Application of Remote Sensing GIS in Agriculture.pdfCrystal Sanchez
 
An Integrated Parametric Approach To Landfill Site Selection Fuzzy GIS-Based ...
An Integrated Parametric Approach To Landfill Site Selection Fuzzy GIS-Based ...An Integrated Parametric Approach To Landfill Site Selection Fuzzy GIS-Based ...
An Integrated Parametric Approach To Landfill Site Selection Fuzzy GIS-Based ...IJERA Editor
 
AGRO 301 Introduction Class.pptx
AGRO 301 Introduction Class.pptxAGRO 301 Introduction Class.pptx
AGRO 301 Introduction Class.pptxArchanaNancy1
 
IRJET - IoT based Fertilizer Injector for Agricultural Plants
IRJET -  	  IoT based Fertilizer Injector for Agricultural PlantsIRJET -  	  IoT based Fertilizer Injector for Agricultural Plants
IRJET - IoT based Fertilizer Injector for Agricultural PlantsIRJET Journal
 
““Smart Crop Prediction System and Farm Monitoring System for Smart Farming””
““Smart Crop Prediction System and Farm Monitoring System for Smart Farming””““Smart Crop Prediction System and Farm Monitoring System for Smart Farming””
““Smart Crop Prediction System and Farm Monitoring System for Smart Farming””IRJET Journal
 
Geographical Information System
Geographical Information SystemGeographical Information System
Geographical Information SystemDevegowda S R
 
GIS Applications for Smart Agriculture-Case Studies & Research Prospects.
GIS Applications for Smart Agriculture-Case Studies & Research Prospects.GIS Applications for Smart Agriculture-Case Studies & Research Prospects.
GIS Applications for Smart Agriculture-Case Studies & Research Prospects.AdityaAllamraju1
 
Precision Farming and Good Agricultural Practices (1).pptx
Precision Farming and Good Agricultural Practices (1).pptxPrecision Farming and Good Agricultural Practices (1).pptx
Precision Farming and Good Agricultural Practices (1).pptxNaveen Prasath
 

Similar to Enhancement PPT.pptx (20)

Applications of Remote Sensing
Applications of Remote SensingApplications of Remote Sensing
Applications of Remote Sensing
 
Machine learning project
Machine learning projectMachine learning project
Machine learning project
 
IRJET - Survey on Soil Classification using Different Techniques
IRJET - Survey on Soil Classification using Different TechniquesIRJET - Survey on Soil Classification using Different Techniques
IRJET - Survey on Soil Classification using Different Techniques
 
Publisher in research
Publisher in researchPublisher in research
Publisher in research
 
Land use Land Cover Highlight for Jibia Local Government, Nigeria
Land use Land Cover Highlight for Jibia Local Government, NigeriaLand use Land Cover Highlight for Jibia Local Government, Nigeria
Land use Land Cover Highlight for Jibia Local Government, Nigeria
 
International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)
 
Q33081091
Q33081091Q33081091
Q33081091
 
Precision agriculture
Precision agriculturePrecision agriculture
Precision agriculture
 
IRJET - Normalized Difference Vegetation Index (NDVI) based Land Cover Cl...
IRJET -  	  Normalized Difference Vegetation Index (NDVI) based Land Cover Cl...IRJET -  	  Normalized Difference Vegetation Index (NDVI) based Land Cover Cl...
IRJET - Normalized Difference Vegetation Index (NDVI) based Land Cover Cl...
 
An automated low cost IoT based Fertilizer Intimation System forsmart agricul...
An automated low cost IoT based Fertilizer Intimation System forsmart agricul...An automated low cost IoT based Fertilizer Intimation System forsmart agricul...
An automated low cost IoT based Fertilizer Intimation System forsmart agricul...
 
Application of Remote Sensing GIS in Agriculture.pdf
Application of Remote Sensing   GIS in Agriculture.pdfApplication of Remote Sensing   GIS in Agriculture.pdf
Application of Remote Sensing GIS in Agriculture.pdf
 
An Integrated Parametric Approach To Landfill Site Selection Fuzzy GIS-Based ...
An Integrated Parametric Approach To Landfill Site Selection Fuzzy GIS-Based ...An Integrated Parametric Approach To Landfill Site Selection Fuzzy GIS-Based ...
An Integrated Parametric Approach To Landfill Site Selection Fuzzy GIS-Based ...
 
AGRO 301 Introduction Class.pptx
AGRO 301 Introduction Class.pptxAGRO 301 Introduction Class.pptx
AGRO 301 Introduction Class.pptx
 
PPT.Geoinformatics.pdf
PPT.Geoinformatics.pdfPPT.Geoinformatics.pdf
PPT.Geoinformatics.pdf
 
IRJET - IoT based Fertilizer Injector for Agricultural Plants
IRJET -  	  IoT based Fertilizer Injector for Agricultural PlantsIRJET -  	  IoT based Fertilizer Injector for Agricultural Plants
IRJET - IoT based Fertilizer Injector for Agricultural Plants
 
““Smart Crop Prediction System and Farm Monitoring System for Smart Farming””
““Smart Crop Prediction System and Farm Monitoring System for Smart Farming””““Smart Crop Prediction System and Farm Monitoring System for Smart Farming””
““Smart Crop Prediction System and Farm Monitoring System for Smart Farming””
 
Geographical Information System
Geographical Information SystemGeographical Information System
Geographical Information System
 
GIS Applications for Smart Agriculture-Case Studies & Research Prospects.
GIS Applications for Smart Agriculture-Case Studies & Research Prospects.GIS Applications for Smart Agriculture-Case Studies & Research Prospects.
GIS Applications for Smart Agriculture-Case Studies & Research Prospects.
 
Assignment agus
Assignment agusAssignment agus
Assignment agus
 
Precision Farming and Good Agricultural Practices (1).pptx
Precision Farming and Good Agricultural Practices (1).pptxPrecision Farming and Good Agricultural Practices (1).pptx
Precision Farming and Good Agricultural Practices (1).pptx
 

Recently uploaded

Porous Ceramics seminar and technical writing
Porous Ceramics seminar and technical writingPorous Ceramics seminar and technical writing
Porous Ceramics seminar and technical writingrakeshbaidya232001
 
VIP Call Girls Service Hitech City Hyderabad Call +91-8250192130
VIP Call Girls Service Hitech City Hyderabad Call +91-8250192130VIP Call Girls Service Hitech City Hyderabad Call +91-8250192130
VIP Call Girls Service Hitech City Hyderabad Call +91-8250192130Suhani Kapoor
 
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptxDecoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptxJoão Esperancinha
 
High Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur EscortsHigh Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur Escortsranjana rawat
 
MANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLS
MANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLSMANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLS
MANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLSSIVASHANKAR N
 
Introduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptxIntroduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptxupamatechverse
 
Analog to Digital and Digital to Analog Converter
Analog to Digital and Digital to Analog ConverterAnalog to Digital and Digital to Analog Converter
Analog to Digital and Digital to Analog ConverterAbhinavSharma374939
 
Introduction to Multiple Access Protocol.pptx
Introduction to Multiple Access Protocol.pptxIntroduction to Multiple Access Protocol.pptx
Introduction to Multiple Access Protocol.pptxupamatechverse
 
The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...
The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...
The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...ranjana rawat
 
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur Escorts
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur EscortsCall Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur Escorts
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur EscortsCall Girls in Nagpur High Profile
 
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICS
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICSHARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICS
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICSRajkumarAkumalla
 
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...Dr.Costas Sachpazis
 
Extrusion Processes and Their Limitations
Extrusion Processes and Their LimitationsExtrusion Processes and Their Limitations
Extrusion Processes and Their Limitations120cr0395
 
Introduction to IEEE STANDARDS and its different types.pptx
Introduction to IEEE STANDARDS and its different types.pptxIntroduction to IEEE STANDARDS and its different types.pptx
Introduction to IEEE STANDARDS and its different types.pptxupamatechverse
 
(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts
(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts
(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escortsranjana rawat
 
Processing & Properties of Floor and Wall Tiles.pptx
Processing & Properties of Floor and Wall Tiles.pptxProcessing & Properties of Floor and Wall Tiles.pptx
Processing & Properties of Floor and Wall Tiles.pptxpranjaldaimarysona
 
VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130
VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130
VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130Suhani Kapoor
 
Call Girls Delhi {Jodhpur} 9711199012 high profile service
Call Girls Delhi {Jodhpur} 9711199012 high profile serviceCall Girls Delhi {Jodhpur} 9711199012 high profile service
Call Girls Delhi {Jodhpur} 9711199012 high profile servicerehmti665
 

Recently uploaded (20)

Porous Ceramics seminar and technical writing
Porous Ceramics seminar and technical writingPorous Ceramics seminar and technical writing
Porous Ceramics seminar and technical writing
 
VIP Call Girls Service Hitech City Hyderabad Call +91-8250192130
VIP Call Girls Service Hitech City Hyderabad Call +91-8250192130VIP Call Girls Service Hitech City Hyderabad Call +91-8250192130
VIP Call Girls Service Hitech City Hyderabad Call +91-8250192130
 
★ CALL US 9953330565 ( HOT Young Call Girls In Badarpur delhi NCR
★ CALL US 9953330565 ( HOT Young Call Girls In Badarpur delhi NCR★ CALL US 9953330565 ( HOT Young Call Girls In Badarpur delhi NCR
★ CALL US 9953330565 ( HOT Young Call Girls In Badarpur delhi NCR
 
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptxDecoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
 
High Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur EscortsHigh Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur Escorts
 
MANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLS
MANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLSMANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLS
MANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLS
 
Introduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptxIntroduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptx
 
Analog to Digital and Digital to Analog Converter
Analog to Digital and Digital to Analog ConverterAnalog to Digital and Digital to Analog Converter
Analog to Digital and Digital to Analog Converter
 
Introduction to Multiple Access Protocol.pptx
Introduction to Multiple Access Protocol.pptxIntroduction to Multiple Access Protocol.pptx
Introduction to Multiple Access Protocol.pptx
 
The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...
The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...
The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...
 
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur Escorts
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur EscortsCall Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur Escorts
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur Escorts
 
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICS
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICSHARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICS
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICS
 
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
 
Extrusion Processes and Their Limitations
Extrusion Processes and Their LimitationsExtrusion Processes and Their Limitations
Extrusion Processes and Their Limitations
 
Introduction to IEEE STANDARDS and its different types.pptx
Introduction to IEEE STANDARDS and its different types.pptxIntroduction to IEEE STANDARDS and its different types.pptx
Introduction to IEEE STANDARDS and its different types.pptx
 
(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts
(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts
(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts
 
Processing & Properties of Floor and Wall Tiles.pptx
Processing & Properties of Floor and Wall Tiles.pptxProcessing & Properties of Floor and Wall Tiles.pptx
Processing & Properties of Floor and Wall Tiles.pptx
 
VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130
VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130
VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130
 
Roadmap to Membership of RICS - Pathways and Routes
Roadmap to Membership of RICS - Pathways and RoutesRoadmap to Membership of RICS - Pathways and Routes
Roadmap to Membership of RICS - Pathways and Routes
 
Call Girls Delhi {Jodhpur} 9711199012 high profile service
Call Girls Delhi {Jodhpur} 9711199012 high profile serviceCall Girls Delhi {Jodhpur} 9711199012 high profile service
Call Girls Delhi {Jodhpur} 9711199012 high profile service
 

Enhancement PPT.pptx

  • 1. Presented by D N Kiran Pandiri Reg. No. 19-3-04-118 Supervisor Dr. R Murugan, Assistant Professor Department of Electronics and Communication Engineering National Institute of Technology, Silchar Soil and Crop Leaf Disease Classification using Artificial Intelligence Proposed Tentative Title
  • 2. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 2 I Welcome the Doctoral Committee to the Enhancement Seminar  Chairman : Dr. T Khan, Asst Prof, ECE  External Examiner : Dr. Sumantra Dutta Roy, Professor, EE, IIT Delhi  Member : Dr. Chandrajit Choudhury, Asst Prof, ECE  Nominee of Chairman, Senate : Dr. Asha Rani M A, Asst Prof, EE  Supervisor : Dr. R Murugan, Asst Prof, ECE
  • 3. Outline of the presentation  Broad area of research  Introduction  Literature survey  Problem statements  Objectives  Conclusion  Publications  Works under communication  References 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 3
  • 4. Broad area of research  Image Processing in Agriculture 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 4
  • 5. Introduction  It is a practice of cultivating plants and livestock.  It has been practicing from many centuries by our ancestors and by us till now.  Agriculture with its allied sectors is the largest source of livelihood in many countries across the globe. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 5 Fig.1. Image representation of cultivation and livestocks. [1]
  • 6. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 6 The agriculture and forest lands decreased for the past two decades [2][3]. The global food demand has been increasing with increase in population. Fig. 2. Agriculture and forest land area declined in statistics from 2000 – 2018 [2]
  • 7. Crop Loss due to diseases 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 7 0 50 100 Yield Loss Between 20 - 40 percent of Global crop yields are reduced each year due to the damage wrought by plant pests and diseases.
  • 8. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 8 Usage of pesticides and fertilizers Fig. 3. Percentage increase in the usage of fertilizers [2]
  • 9. Decrease in work force 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 9 Fig 4. The decline of workforce in the last two decades [2]
  • 10. Digital Transformation of Agriculture  There is a need for optimizing the agriculture practices without putting extra environmental burden.  The digital transformation of the agriculture sector is to meet the increasing food demand of the global population explosion without affecting the environmental conditions and maintaining the profit to the farmers.  Digitalization of agriculture can reduce the burden of employee workforce, crop management, soil management and many other. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 10
  • 11. Motivation  Digitalization of the agriculture sector that reduces the use of fertilizers and pesticides and increases the crop yield by maintaining profits to the farmers.  We have consider two key components which can reduce the usage of fertilizers and pesticides in increasing the crop yield is 1. Soil type classification 2. Plant leaf disease classification 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 11
  • 12. Soil Type Classification  Soil is the primary component in agriculture.  Soil type identification is the key to the success of site-specific crop management.  Soil management helps in varying the tillage [4].  Identifying the soil type can reduce or minimize the soil tillage by reducing the production cost.  Based on the soil type, the soil conditions like soil texture and moisture content can be adjusted, which are important for seed and fertilizer placement.  There is no effective research on physical properties of the soil like soil type, porosity, texture, structure, density, etc., 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 12
  • 13.  The conventional methods used for identifying the soil type are  Laboratory testing [7][8]  In-situ testing [5][6]  All these methods are time-consuming, high cost, and require proficiency to analyze soil type.  By using digital techniques, researchers used texture, color, particle size, pattern, pH level, and organic matter as features in distinguishing the different types of soils.  The soil type is identified based on the texture of the soil sample.  The soil texture varies on the percentage of sand, silt, and clay present in the soil sample. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 13
  • 14. Sample Soil Images 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 14 Fig 5. Sample Soil images. (a) Loamy sand, (b)&(d) sand, (c),(e)&(f) clay, and (g) sandy loam. (a) (b) (c) (d) (e) (f) (g)
  • 15. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 15 [Ref] Year Title Input Data Methods Focus [9]2015 Soil type classification and mapping using hyperspectral remote sensing data. Hyperspectral image Gaussian SVM Classify the soils based on the colour with hyperspectral data [10]2019 Compressive spectral imaging system for soil classification with three-dimensional convolutional neural network. 3D CNN Hyperspectral images are used in classification of soil type. PCA is used in reducing the feature dimensions. [11]2016 A smartphone-based soil color sensor: For soil type classification. Smart Phone Camera with spectral lenses Linear Discriminant Analysis Classified the soil types based on the Munsel color. The color of the soil identified by using the spectral wavelength. [12]2020 Classification of soil aggregates: A novel approach based on deep learning. Stereo pair images CNN Deep Learning methods in classifying aggregate soil [13]2012 Soil texture classification algorithm using RGB characteristics of soil images. CCD Camera Images Linear Regression Using RGB histogram of the CCD camera images, classified the soil types Literature Review on Soil Classification
  • 16. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 16 [Ref] Year Title Input Data Methods Focus [14]2018 Soil Characterization and Classification: A Hybrid Approach of Computer Vision and Sensor Network Camera images and Sensor data NN Features of Soil images were extracted from the images captured using the DSLR camera and moisture level of the soils were measured using a sensor. All these data was used in classification of soil types. [151999 An artificial neural network for classifying and predicting soil moisture and temperature using Levenberg-Marquardt algorithm. Sensor data NN with Levenberg- Marquardt Classified two soil classes grass and bare smooth soils based on the remotely sensed data to measure the moisture of the soil. Measured the moisture levels at different depths for the same soil samples and classify the soil types. [16]2017 Performance of SVM classifier for image based soil classification. Android and SLR Images SVM Classified the soil types by extracting the features using LPF, Gabor filter bank, and color momentum of the soil images obtained from online mobile images [17]2020 Soil texture classification using multi class support vector machine Texture features of the labelled soil samples were extracted by using HSV histogram, color moment, Gabor wavelets, and DWT. The features were classified using SVM classifier. The soil samples were collected from paddy fields in Guwahati. [18]2020 Artificial intelligence system for supporting soil classification. CNN Classified the soils sand, clay and gravel using SLR camera images
  • 17. Research Gaps  Even though many promising methodologies have been developed and implemented, research on classification of soil type using mobile/SLR camera images is in initial stages.  Classification of soil type using mobile/SLR camera images is a challenging task.  There is no publicly available soil image dataset created by following the standards in capturing the images of the soil type.  Classifying the soil type using the texture features of the soil type is complicated but might give accurate results as the RGB characteristics of the soil vary from region to region due to climate and environmental conditions.  Soil classification using images can reduce the computational time and cost in analyzing the soil type.  Soil classification will help in soil management in agriculture sector. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 17
  • 18. Crop Leaf Disease Classification  The crop management is an important task at every stage of the of the crop growth.  The crop management involves Yield Prediction Crop quality Leaf disease identification Weed detection Crop recognition  To meet the necessity of the people, the production of the crop has to be increased.  One of the most important feature in crop management is disease classification.  Early detection of diseases in plants can reduce the pesticides and fertilizers usage in protecting and increasing the yield of the crop. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 18
  • 19. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 19 Fig 6. Percentage of primary crops production across the world. [19].
  • 20. Most investigated crops[29] 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 20 Fig 7. Research article count on the primary crop [19]
  • 21.  Potato (Solanum Tuberosum) has been chosen for plant leaf disease detection.  Potato stands top in most consuming vegetable across the globe.  India stands second largest producer nation, with 53.69 million metrics tons produced during the fiscal year 2021.  The average consumption of potato by a person is 49.4 pounds during the year 2019 [20].  Potato leaves are affected due to fungal pathogens is Early Blight (Alternaria) Late Blight (Phytophthora)  The annual potato crop loss due to late blight alone is 6.7 billion dollars across the world. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 21
  • 22. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 22 Early Blight Late Blight
  • 23. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 23 Literature Review on Potato Leaf Disease Classification [Ref] Year Title Input Data Methods Focus [21]2018 Using Support Vector Machines classification to differentiate spectral signatures of potato plants infected with Potato Virus Y Hyperspectral images SVM Classification of potato virus Y infected plants using unmanned aerial vehicle to capture the hyperspectral images with a spectral range 350-2500 nm in NIR and SWIR wavelengths. The spatial samples were classified using SVM [22]2021 An Empirical Study on Machine Learning Models for Potato Leaf Disease Classification using RGB Images Central Potato Research Institute (CPRI), India, Plantvillage Gabor filter, Fine tuned VGG16,SVM Collected the real time images from the field and segmented the region of interest and the features are extracted using FCNN which are used in ML for classification using SVM. [23]2021 Automated abnormal potato plant detection system using deep learning models and portable video cameras Video frame images CNN Abnormal potato plants were identified during the early growth and mid-growth of the plant [24]2020 Detection of Potato Disease Using Image Segmentation and Multi class Support Vector Machine Plantvillage GLCM, Hu, Color Histogram, ML Feature extraction from the diseased segmented part and classifying the leaves
  • 24. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 24 [Ref] Year Title Input Data Methods Focus [25]2021 Ensemble Classification and Feature Extraction Based Plant Leaf Disease Recognition Plantvillage Dataset GLCM+LBP+PCA +LAW texture mask+LDA+ML Ensemble methods are created for extracting the features using different sizes convolution mask and the features are used in classifying the leaf disease images of Tomato, Bell Pepper and Potato [26]2021 Recognition of early blight and late blight diseases on potato leaves based on graph cut segmentation Graph cut + LBP + ML (SVM) For Classifying the potato leaf diseases at different stages, dataset has been segregated and segmented the images using Graph cut model to detect the diseases. The features of the images are extracted by using LBP and fed to ML classifier algorithms. [27]2021 Automated plant leaf disease detection and classification using optimal MobileNet based convolutional neural networks Extreme learning + ML Classified the tomato plant leaf diseases by optimizing the hyperparameters of the model using EPO algorithm [28]2022 Transfer Learning for Multi-Crop Leaf Disease Image Classification using Convolutional Neural Network VGG VGG16 Multi crop leaf disease classification at early stages in tomato and grape crop leaf. The diseased spot is annotated by rectangular bounding box. [29]2020 Identification of Disease in Potato Leaves Using Convolutional Neural Network (CNN) Algorithm CNN Classification of potato leaf diseases Early Blight and Late Blight
  • 25. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 25 Research Gap  Research in identifying the diseased leaves in potato plants is still in early stages.  Classification is carrying on the laboratory images for developing the models.  Segmentation of infected leaf in identifying the diseases is a challenging task.  Availability of real time images is low.  There is no cost effective infrastructure in monitoring the crop disease leaves.
  • 26. Problem Statement 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 26 1. Creating the standard soil image database. 2. Classifying the standard soil types sand and clay using machine learning algorithm 3. Multi-class soil classification using the soil image database created by considering the imbalance data. 4. Classification of Phytophthora infestans and Alternaria solani diseases in Potato plant leaves using deep learning networks. 5. Detection of diseases caused by fungi pathogens in Solanum Tuberosum leaves.
  • 27. 20-05-2023 Department of ECE, NIT Silchar 27 Objective 1 The standard soil image database has been created by capturing the images of the soil sample using android/SLR camera. Classifying the standard soil types sand and clay using machine learning algorithm Objective 2 Multi-Feature Fusion for Soil Image Feature Extraction and Classification using Machine Learning Multi-class soil classification using the soil image database created by considering the imbalance data Objective 3 A Smart Soil Image Classification System using Lightweight Convolutional Neural Network Classification of Phytophthora infestans and Alternaria solani diseases in Potato plant leaves using deep learning network Objective 4 POT-Net: Solanum Tuberosum (Potato) Leaves Diseases Classification using an Optimized Deep Convolutional Neural Network Detection of diseases caused by fungi pathogens in Solanum Tuberosum leaves Objective 5 Early Blight and Late Blight disease detection using deep learning models Creating the standard soil image database
  • 28. Problem Statement 1 Objective  The standard soil image database has been created by capturing the images of the soil sample using android/SLR camera. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 28
  • 29. Acquiring Soil Samples  The soil samples are collected from different regions of Andhra Pradesh and Silchar, Assam.  A total of 96 soil samples are gathered from agricultural fields in these locations.  The regions include the Godavari river coastal zone, rich in alluvial and sandy soils, drought- prone areas with sandy and rocky soils, and hill stations suitable for plantation.  The soil samples were obtained at a depth of 5 to 10 cm below the field surface. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 29 (a) (b) (c) Fig. 8. Represents the areas where soil samples are collected. (a) Rajahmundry (b) Madanapalle (c) Digged area
  • 30. Soil analysis in the laboratory 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 30  Each soil sample is subjected to laboratory examination to determine the texture of the collected soil samples and label the soil type by identifying the fraction of sand, silt, and clay.  Sieve Analysis  Hydrometer test • In sieve analysis, 500gms of sand type soils are taken and sieved through the different sieve pans of sieve size up to 75μm to find the fraction of sand. If the fraction of silt and clay that passed through the 75μm sieve pan is greater than 10 percent, we go for a hydrometer test to find the silt and clay fraction. Fig.9. Sieve shaker for Sieve Analysis
  • 31.  For hydrometer test, 33gms of sodium hexametaphosphate and 7gms of sodium bicarbonate in one liter of water, a solution is made and left for 24 hours.  Now, 100ml of the solution is mixed with 800ml of water for the hydrometer test.  In the 900ml prepared solution, 50gm of the tested soil sample is added, and readings are taken at regular intervals over the following 24 hours.  The height of the hydrometer from the surface of the solution is recorded at regular intervals. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 31 • The texture of the soil sample is determined by plotting the percentage of sand, silt and clay on the United States Department of Agriculture (USDA) texture triangle. (a) (b) Fig. 10. (a) Hydrometer and (b) Hydrometer test
  • 32. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 32 (a) (b) Fig 11. a) USDA texture triangle for classification. b) USDA texture triangle with the some soil samples plotted.
  • 33. Experimental Setup  The images of the acquired soil samples are captured by using a Samsung smartphone camera.  The organic components present in the soil are removed, and the moisture content is removed by drying the soil in a hot air oven at 1000C for 24 hours.  A chamber with dimensions 45x20x10 cm was constructed with Styrofoam to capture the images. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 33 • As the box walls are white, allowing the light that enters the box through the window will provide the illumination conditions enough to capture the soil images without affecting the actual color of the soil. (a) (b) Fig 12. Image acquisition setup. a) Isometric view of a chamber (b) Styrofoam box to capture images.
  • 34.  The soil images are captured using a smartphone camera with 48 megapixels.  A total of 392 soil images were captured from the collected samples.  The dataset is made available in the IEEE data portal. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 34 (a) (b) (c) Fig 13. Represents the clay, loamy sand, and sandy loam soil images captured using a smartphone camera.
  • 35. Summary  This work presents the creation soil image database.  The soil image database created by following standards so that the soil samples can be labelled accurately.  The database may help the researchers in identifying the soil types by image analysis.  This can enable in advancement of agriculture towards automation. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 35
  • 36. Objective  Multi-Feature Fusion for Soil Image Feature Extraction and Classification using Machine Learning 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 36 Problem Statement 2
  • 37. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 37 • Contributions • Initially 70 soil samples were collected from different regions of Andhra Pradesh • The feature fusion technique has been proposed to extract the soil texture features. • The GSVM machine learning classifier has been trained to classify the soil types by varying the kernel parameter. (a) (b) (c) Fig. 14. ROI extracted images (a) Coarse sand (b) Fine sand (c) Clay soil images. Fig.15. Work Flow of the model
  • 38. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 38 𝐺𝐿𝐶𝑀 = 𝑝 1,1 𝑝 1,2 ⋯ 𝑝 1, Mg 𝑝 2,1 𝑝 2,2 ⋯ 𝑝 2, Mg ⋮ ⋮ ⋱ ⋮ 𝑝 Mg, 1 𝑝 Mg, 2 ⋯ 𝑝 Mg, Mg . . 1 • Calculate the features: contrast, correlation, energy, homogeneity, and entropy using GLCM [30]. • Calculate the Tamura features [31]. • Gabor filter bank is created by using two scales, six orientations, and four frequency values [32]. • The ROI images are quantized with hue, saturation, and value with 6 × 3 × 5 equal bins. The output is a vector of size 1 × 90. • The feature fusion vector size for each image is 1 × 202. (a) (b) (c) (d) Fig 16. a) Confusion matrix of the proposed model. (b) True positive and False-negative rate of the model. (c) ROC curve of clay and (d) ROC curve of sand.
  • 39. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 39 (a) (b) (c) (d) (e) (f) Fig.17. Confusion matrix of GSVM using (a) GLCM (b) Tamura (c) T+GLCM (d) Gabor (e) Gabor+GLCM (f) Feature fusion model State-of- the-art GLCM Tamura T+ GLCM Gabor Gabor + GLCM Feature Fusion GSVM [33] 85 82.5 80 85 95 97.5 Table 1. Ablation experimental analysis State-of- the-art GLCM Tamura T+ GLCM Gabor Gabor+ GLCM Feature Fusion LR [34] 65 60 90 72.5 62.5 57.5 NB [36] 55 55 57.5 57.5 62.5 70 WKNN [35] 90 62.5 77.5 77.5 90 90 BNN [37] 95 65 72.5 72.5 70 92.5 SVM 42.5 60 52.5 67.5 62.5 87.5 GSVM 85 82.5 80 85 95 97.5 Table 2. Comparative analysis of the state-of-the-art in texture features and ML classification algorithms
  • 40. Algorithm Precision Recall Specificity F1-score Error rate Accuracy (%) LR 0.55 0.578 0.571 0.564 0.85 57.5 NB 0.8 0.67 0.75 0.727 0.6 70 GSVM 0.95 1 0.952 0.974 0.05 97.5 WKNN 0.9 0.9 0.9 0.9 0.1 90 SVM 0.9 0.857 0.895 0.878 0.125 87.5 B-NN 1 0.869 1 0.929 0.15 92.5 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 40 Fig. 18. Confusion Matrix of (a) LR (b) NB (c) SVM (d) WKNN (e) B-NN (f) GSVM. (a) (b) (c) (d) (e) (f) Table 3. Performance analysis in classifying the soils with state-of-the-art ML algorithms. Fig. 19. Comparison of the proposed model with state-of-the-art using 10-fold cross- validation
  • 41. Summary  This work presents multi-feature fusion model for the classification of standard soils sand and clay using Gaussian SVM.  The feature fusion model is compared with state-of-the-art machine learning classifiers.  This can enable the researchers in advancement and upgrade of techniques for the classification various soil types. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 41
  • 42. Objective Smart Soil Image Classification System using Lightweight Convolutional Neural Network 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 42 Problem Statement 3
  • 43. Contributions  Rather than classifying the gravel and aggregate soils, classified the soil types like sand, clay, loam, sandy loam and loamy sand related to agriculture soils.  HSV, RGB extraction, and adaptive histogram techniques are applied to the database to highlight the texture features of the soil sample images.  Developing a novel Lightweight network aimed to classify the soil images with less number of layers, learnable parameters, epochs, size of the network, and better performance. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 43
  • 44. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 44 Fig. 20. The workflow of the proposed model in the classification of soil.
  • 45. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 45 (a) (b) (c) Fig. 21. a, b, and c represent the clay, loamy sand and sand images and the corresponding ROI images. (a) (b) (c) (d) (e) Fig. 22. Sample Pre-processed images. (a) Adaptive histogram [38]. (b-d) RGB channels respectively. [39] (e) V extraction from HSV.
  • 46. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 46 Fig. 23. Proposed Light-SoilNet network architecture
  • 47. Class TP TN FP FN Precision Recall F1 score Clay 146 430 0 2 1.000 0.986 0.993 Loam 90 478 10 0 0.900 1.000 0.947 Sandy Loam 100 465 0 13 1.000 0.885 0.939 Loamy Sand 121 454 5 2 0.960 0.984 0.972 Sand 104 476 0 2 1.000 0.981 0.990 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 47 Adaptive Histogram V value RGB Channels Accuracy F1 Score Clay Loam Loamysand Sand Sandyloam -   81.4 0.83 0.92 0.95 0.64 0.82  -  78.4 0.79 0.50 0.92 0.86 0.55   - 92.8 0.94 0.75 0.88 0.95 1.00    97.2 0.99 0.94 0.97 0.99 0.94 Table 5. Ablation experiment results for preprocessed techniques. Fig. 24. Confusion matrix and ROC of the Light-SoilNet network. Table 4. Performance metrics of the proposed model for each class.
  • 48. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 48 Architecture Accuracy (%) Size (MB) Proposed model 97.20 9.62 MobileNet-v2 [40] 95.45 21.5 MobileNet-v3 94.5 21.1 EfficientNet-B0 [41] 87.54 28.3 ShuffleNet [42] 95.67 14.5 Network Clay Loam Sand Sandy loam Loamy sand P R P R P R P R P R MobileNet-v2 0.918 0.944 1 0.980 0.885 0.920 1 1 1 0.940 ShuffleNet 0.973 0.986 1 1 0.906 1.000 1 0.901 0.913 0.906 EfficientNet- B0 0.849 0.756 0.500 1 1 0.779 1 1 1 0.984 Light-SoilNet 1 0.986 0.900 1 1 0.981 1 0.885 0.960 0.984 Table 6. Classification accuracy of Light-SoilNet and pre-trained lightweight networks. (a) (b) (c) (d) Fig. 25. Confusion matrix. (a) EfficientNet-B0 (b) MobileNet-v2 (c) ShuffleNet (d) MobileNet-v3. Table 7. Performance analysis for five soil classes with pre-trained lightweight networks.
  • 49. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 49 Architecture Accuracy (%) Parameters (Millions) FLOPs (G) Inference Time (sec) Training Time (min) Size (MB) Proposed model 97.20 2.7 3.74 11 16 9.62 Inception-v3 [44] 94.46 24 5.72 6.5 6.5 94.1 Vgg-19 [43] 85.12 144 15.47 7 7.1 519 ResNet-50 [43] 94.46 23 3.87 4.5 6 100 AlexNet [45] 91.00 61 7.27 4.1 7.4 226 Network Clay Loam Sand Sandy loam Loamy sand P R P R P R P R P R AlexNet 0.822 0.923 0.900 1 0.981 0.754 1.000 1 0.889 0.982 Inception 0.918 1 1 1 1 0.779 0.800 1 1 0.984 Vgg-19 0.959 0.833 1 1 0.528 0.651 0.700 1 1 0.685 ResNet-50 1 0.820 1 1 0.962 1 1 1 0.778 1.000 Light- SoilNet 1 0.986 0.900 1 1 0.981 1 0.885 0.960 0.984 Table 8. Classification accuracy with comparative results. (a) (b) (c) (d) Fig. 26. Confusion matrix. (a)VGG-19 (b)AlexNet (c) ResNet-50 (d) Inception-v3. Table 9. Performance analysis for the five soil classes with pre-trained DL networks.
  • 50. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 50 Fig. 27. Comparative analysis of F1-score for the soil images. Fig. 28. Comparison of Light-SoilNet with the pre-trained DL and lightweight networks using accuracy as cross-validation.
  • 51. Model IRSID Dataset Accuracy (%) Online Dataset Accuracy (%) (Azizi, 2020)ResNet-50 94.46 68.1 (Azizi, 2020)Vgg-16 92.8 57.4 (Azizi, 2020)Inception-v3 94.46 68.09 (Inazumi, Ph, 2020) 92.72 70.2 Proposed (Light-SoilNet) 97.2 65.7 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 51 Table 10. Comparison of the Light-SoilNet and IRSID dataset with existing models and online dataset. Summary • This paper presents a new lightweight CNN-based network architecture for classifying the five different types of soil images. • The performance of the model compared with pre-trained lightweight and DL pre-trained networks. • The proposed Light-SoilNet architecture outperforms the state-of- the-art models in classifying the soil images in terms of accuracy, reducing the learnable parameters, FLOPs, and memory of the architecture. • This can enable the researchers in advancement and upgrade of techniques for the classification various soil types using standard images.
  • 52. Objective  POT-Net: Solanum Tuberosum (Potato) Leaves Diseases Classification using an Optimized Deep Convolutional Neural Network 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 52 Problem Statement 3
  • 53. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 53 Fig 29. Schematic diagram of the proposed POT-Net model
  • 54. Size/Operation Activation Filter Depth Stride Number of Parameters Input (224×224×3) - - - - 0 CL_1 224×224×3 7×7 128 1 18944 BNL_1+ReLU_1 224×224×128 - - - 256 Max Pool_1 112×112×128 2×2 - 2 0 CL_2 112×112×128 5×5 128 1 409728 BNL_2+ReLU_2 112×112×128 - - - 256 Max Pool_2 56×56×128 2×2 - 2 0 CL_3 56×56×64 5×5 64 1 204864 BNL_3+ReLU_3 56×56×64 - - - 128 Max Pool_3 28×28×64 2×2 - 2 0 CL_4 28×28×64 3×3 64 1 36928 BNL_4+ReLU_4 28×28×64 - - - 128 Max Pool_4 14×14×64 2×2 - 2 0 CL_5 14×14×32 3×3 32 1 18464 BNL_5+ReLU_5 14×14×32 - - - 64 Max Pool_5 7×7×32 2×2 - 2 0 CL_6 7×7×16 3×3 16 1 4624 BNL_6+ReLU_6 7×7×16 - - - 32 Max Pool_6 3×3×16 2×2 - 2 0 CL_7 3×3×16 3×3 16 1 2320 BNL_7+ReLU_7 3×3×16 - - - 32 Max Pool_7 1×1×16 2×2 - 2 0 CL_8 1×1×8 3×3 8 1 1160 BNL_8+ReLU_8 1×1×8 - - - 16 FC 1×1×3 - - - 27 Softmax 1×1×3 - - - 0 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 54 Contribution • A novel Optimized deep learning network to classify Potato leaf diseases using CNN architecture (POT-Net) has been developed to identify the EB and LB diseased leaves of the potato plant from healthy leaves with less run-time. • The proposed model has an optimum number of parameters compared with the state-of-the-art DL techniques, which reduces the cost of the model. • The hyperparameters of the model are optimized using a metaheuristic algorithm called Whale Optimization Algorithm by improving the efficiency of the model in identifying the diseased leaves with an accuracy of 99.12%. • The proposed model performance has been compared with the state-of-the-art models, pre-trained DL networks and various meta- heuristic optimization algorithms to prove the efficiency of the model. Table 11. Proposed CNN Architecture with parameters and filter values at each layer.
  • 55.  Encircling the prey: 𝐷 = 𝐶. 𝑋𝑟 𝑡 − 𝑋(𝑡) … (7) 𝑋 𝑡 + 1 = 𝑋𝑟 𝑡 − 𝐴 ∙ 𝐷 … (8)  If 𝐴 ≥ 1, then the search agent is far from the optimal solution, and a new search agent is chosen.  . If 𝐴 < 1 best solution is achieved by the search agent.  Exploitation Phase: 𝑋 𝑡 + 1 = 𝑋𝑖 𝑡 − 𝐴 ∙ 𝐷, 𝑖𝑓 𝑝 < 0.5 (𝑆ℎ𝑟𝑖𝑛𝑘𝑖𝑛𝑔 𝑒𝑛𝑐𝑖𝑟𝑙𝑒) 𝐷′ ∙ 𝑒𝑠𝑙 ∙ cos 2𝜋𝑙 + 𝑋𝑖 𝑡 , 𝑖𝑓 𝑝 ≥ 0.5 (𝑆𝑝𝑖𝑟𝑎𝑙 𝑢𝑝𝑑𝑎𝑡𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛) … (12) 𝐷 ′ = 𝑋𝑖 𝑡 − 𝑋(𝑡) 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 55 Whale Optimization Algorithm (WOA) [46] Fig. 30. Creating spiral shape bubbles by shrinking the circles to catch the prey.
  • 56. Model Momentum Epochs L2regularization Learning rate CNN optimized with WOA 0.7070 6.3717 0.0002 0.0892 CNN without optimization 0.01 8 0.0001 0.01 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 56 Class TP TN FP FN Precision Recall F1 score Early Blight 1056 1772 14 4 0.996 0.987 0.991 Healthy 730 2109 7 0 1 0.99 0.995 Late Blight 1035 1786 4 21 0.980 0.996 0.988 (a) (b) (c) Fig 31. Sample images. (a) Healthy (b) Early Blight (c) Late Blight [47] Category No. of images Training set Validation set Early Blight 2472 1060 Healthy 1702 730 Late Blight 2465 1056 Total 6639 2846 Table 12. Dataset Description for training and validation. (a) (b) Results Fig 32. POT-Net model (a) Confusion matrix (b) ROC Table 14. POT-Net model performance parameters and metrics. Table 13. Training options of the CNN with and without optimization Model Accuracy (%) Learnable Parameters (millions) Computation time per image (sec) Sholihati et al. [48] 91.31 22 0.34 Deepa et al. [49] 88 - 0.6 Network1 93.64 0.3 0.08 Network2 97.26 1.56 0.15 POT-Net 99.12 0.7 0.12 Table 15. Performance analysis with computational time and learnable parameters.
  • 57. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 57 Architecture Accuracy (%) Parameters (Millions) AlexNet [45] 97.79 61 DenseNet [33] 96.21 7.2 ResNet-50 [43] 91.64 23 MobileNet-v2 [40] 90.55 3.47 ShuffleNet [42] 98.63 2.29 EfficientNet-B0 [41] 98.77 26 VGG-19 [43] 98.7 144 Non-Opt CNN 97 0.7 POT-Net 99.12 0.7 Network Early Blight Healthy Late Blight Precision Recall F1-score Precision Recall F1-score Precision Recall F1- score AlexNet 0.988 0.982 0.985 0.953 0.998 0.975 0.985 0.96 0.972 DenseNet 0.969 0.986 0.977 0.957 0.955 0.956 0.959 0.941 0.949 ResNet-50 0.968 0.934 0.951 0.898 0.915 0.906 0.875 0.898 0.886 MobileNet 0.875 0.971 0.921 0.993 0.839 0.909 0.876 0.899 0.887 ShuffleNet 0.982 0.992 0.986 0.986 0.995 0.99 0.991 0.974 0.982 EfficientNet-B0 0.982 0.986 0.984 0.997 0.998 0.997 0.986 0.981 0.983 VGG-19 0.981 0.992 0.986 0.991 0.991 0.991 0.989 0.978 0.983 Non-Opt CNN 0.974 0.995 0.984 1 0.921 0.958 0.947 0.985 0.965 POT-Net 0.996 0.987 0.991 1 0.99 0.995 0.98 0.996 0.988 Performance analysis with pre-trained DL and non-optimized CNN network Table 16. Classification accuracy of POT-Net and pre-trained networks. Table 17. Performance metrics of the pre-trained DL networks.
  • 58. Optimizer Accuracy (%) GWO [39] 98.35 SMA [40] 98.91 PSO [41] 97.29 POT-Net 99.12 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 58 Performance analysis with MHO algorithms Optimizer Early Blight Healthy Late Blight Precision Recall F1-score Precision Recall F1-score Precision Recall F1-score GWO 0.983 0.988 0.985 0.981 0.994 0.988 0.986 0.972 0.979 SMA 0.987 0.989 0.988 1.000 0.992 0.996 0.984 0.988 0.986 PSO 0.973 0.982 0.977 0.996 0.953 0.974 0.957 0.979 0.968 POT-Net 0.996 0.987 0.991 1 0.99 0.995 0.98 0.996 0.988 Table 18. Optimizer algorithms performance with proposed CNN architecture. Table 19. Performance metrics of the proposed CNN with the optimizer algorithms. (a) (b) (c) Fig. 33.. Confusion matrix. (a) GWO (b) SMA (c) PSO.
  • 59. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 59 (a) (b) Fig. 34. Testing the POT-Net model with real-time images. (a) Confusion matrix (b) ROC The proposed model has been validated with real-time images containing 395 images. The model has identified the EB, LB and healthy leaves images with an accuracy of 81%.
  • 60. Summary  This paper uses image phenotyping to classify the potato plant leaves when infected with diseases.  An image-based framework for identifying the fungal pathogen disease leaves using a CNN architecture has been presented in this article.  The proposed POT-Net model performance is evaluated with accuracy, precision, recall, and F1-score and compared with the pre-trained DL networks and state-of-the-art optimization algorithms.  The model has achieved high performance in classifying the accuracy is 99.12%, and the time to process an image to classify the diseased leaves is 0.12 sec.  In future, by exploring different CNN architectures, the performance will be extended in identifying the leaf diseases of potato and other plant leaf diseases using laboratory and field images.  the proposed methodology might be viewed as an effective method for the early detection of diseased and healthy leaves in potato plants in minimizing the significant agriculture loss, which helps in automation and continuous monitoring of the fields. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 60
  • 61.  Digital transformation in Agriculture can meet the food demand of the global population with less workforce.  Soil type classification can increase the crop yield by choosing the site specific crops and automation of the agriculture sector.  Standard soil image database has been created to classify the soil types using various machine learning and deep learning approaches by increasing the classification accuracy and minimizing the complexity of the model.  To increase the yield of the crop, detecting the plant fungal diseases at early stages was achieved by using deep learning approaches.  The performance of the proposed models are compared with other state-of-the-art techniques. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 61 Conclusion
  • 62. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 62 Publications Conferences 1. D. N. K. Pandiri, R. Murugan and T. Goel, "ODNet: Optimized Deep Convolutional Neural Network for Classification of Solanum Tuberosum Leaves Diseases," 2022 IEEE Region 10 Symposium (TENSYMP), Mumbai, India, 2022, pp. 1-6, doi: 10.1109/TENSYMP54529.2022.9864335 Database 1. D N Kiran Pandiri, R Murugan, Tripti Goel, March 18, 2021, "Indian Regions Soil Image Database (IRSID): A dataset for classification of Indian soils", IEEE Dataport, doi: https://dx.doi.org/10.21227/2zz3-f173 Journals 1. D. N. Kiran Pandiri, R. Murugan, Tripti Goel, Nishant Sharma, Aditya Kumar Singh, Soumya Sen & Tonmoy Baruah (2023) POT-Net: solanum tuberosum (Potato) leaves diseases classification using an optimized deep convolutional neural network, The Imaging Science Journal, DOI: 10.1080/13682199.2023.2169988 (SCIE, I.F – 0.871)
  • 63. Works communicated 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 63 Journals 1. D N Kiran Pandiri, R. Murugan, T. Goel “Multi-Feature Fusion for Soil Image Feature Extraction and Classification using Machine Learning,” Signal Image and Video Processing (under Review). 2. D N Kiran Pandiri, R. Murugan, T. Goel “Smart Soil Image Classification System using Lightweight Convolutional Neural Network,” Expert Systems with Applications (Under Review). 3. D N Kiran Pandiri, R.Murugan, T. Goel “Classification of Soil using Machine Learning and Deep Learning Techniques–A Review,” IEEE Access (Under Review).
  • 64. References 1. https://www.downtoearth.org.in/factsheet/how-livestock-farming-affects-the-environment-64218 2. https://www.fao.org/3/cb1329en/online/cb1329en.html#chapter-1 3. FAO. 2020. World Food and Agriculture - Statistical Yearbook 2020. Rome. https://doi.org/10.4060/cb1329en 4. https://www.sciencedirect.com/topics/earth-and-planetary-sciences/soil-management 5. Rajapakse, R., 2015. Geotechnical Engineering Calculations and Rules of Thumb: Second Edition, Geotechnical Engineering Calculations and Rules of Thumb: Second Edition. Butterworth-Heinemann. 6. Meigh, A.C., 2013. Cone Penetration Testing: Methods and Interpretation. Elsevier. 7. Bouyoucos, G.J., 1962. Hydrometer Method Improved for Making Particle Size Analyses of Soils 1 . Agron. J. 54, 464–465. https://doi.org/10.2134/agronj1962.00021962005400050028x 8. Lu, N., Ristow, G.H., Likos, W.J., 2000. The Accuracy of Hydrometer Analysis for Fine-Grained Clay Particles. Geotech. Test. J. 23, 487–495. https://doi.org/10.1520/gtj11069j 9. Vibhute, A.D., Kale, K.V., Dhumal, R.K. and Mehrotra, S.C., 2015, December. Soil type classification and mapping using hyperspectral remote sensing data. In 2015 International Conference on Man and Machine Interfacing (MAMI) (pp. 1-4). IEEE.Vibhute, A.D., Kale, K.V., Dhumal, R.K. and Mehrotra, S.C., 2015, December. Soil type classification and mapping using hyperspectral remote sensing data. In 2015 International Conference on Man and Machine Interfacing (MAMI) (pp. 1-4). IEEE. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 64
  • 65. 10. Yu, Y., Xu, T., Shen, Z., Zhang, Y., Wang, X., 2019. Compressive spectral imaging system for soil classification with three- dimensional convolutional neural network. Opt. Express 27, 23029. https://doi.org/10.1364/oe.27.023029 11. Mengistu, A.D. and Alemayehu, D.M., 2018. Soil Characterization and Classification: A Hybrid Approach of Computer Vision and Sensor Network. International Journal of Electrical & Computer Engineering (2088-8708), 8(2). 12. Azizi, A., Gilandeh, Y.A., Mesri-Gundoshmian, T., Saleh-Bigdeli, A.A. and Moghaddam, H.A., 2020. Classification of soil aggregates: A novel approach based on deep learning. Soil and Tillage Research, 199, p.104586. 13. Chung, S.O., Cho, K.H., Cho, J.W., Jung, K.Y., Yamakawa, T., 2012. Soil texture classification algorithm using RGB characteristics of soil images. J. Fac. Agric. Kyushu Univ. 57, 393–397. https://doi.org/10.3182/20101206-3-jp-3009.00005 14. Mengistu, A.D. and Alemayehu, D.M., 2018. Soil Characterization and Classification: A Hybrid Approach of Computer Vision and Sensor Network. International Journal of Electrical & Computer Engineering (2088-8708), 8(2). 15. Atluri, V., Hung, C.C. and Coleman, T.L., 1999, March. An artificial neural network for classifying and predicting soil moisture and temperature using Levenberg-Marquardt algorithm. In Proceedings IEEE Southeastcon'99. Technology on the Brink of 2000 (Cat. No. 99CH36300) (pp. 10-13). IEEE. 16. Srunitha, K., Padmavathi, S., 2017. Performance of SVM classifier for image based soil classification. Int. Conf. Signal Process. Commun. Power Embed. Syst. SCOPES 2016 - Proc. 411–415. https://doi.org/10.1109/SCOPES.2016.7955863 17. Barman, U. and Choudhury, R.D., 2020. Soil texture classification using multi class support vector machine. Information processing in agriculture, 7(2), pp.318-332. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 65
  • 66. 18. Inazumi, S., Ph, D., Intui, S., Eng, M., Jotisankasa, A., Ph, D., Chaiprakaikeow, S., Ph, D., Kojima, K., 2020. Artificial intelligence system for supporting soil classification. Results Eng. 8. https://doi.org/10.1016/j.rineng.2020.100188 19. Benos, L., Tagarakis, A.C., Dolias, G., Berruto, R., Kateris, D. and Bochtis, D., 2021. Machine learning in agriculture: A comprehensive updated review. Sensors, 21(11), p.3758. 20. https://www.ers.usda.gov/data-products/chart-gallery/gallery/chart-detail/?chartId=58340 21. Griffel, L.M., Delparte, D. and Edwards, J., 2018. Using Support Vector Machines classification to differentiate spectral signatures of potato plants infected with Potato Virus Y. Computers and electronics in agriculture, 153, pp.318-324. 22. Ghosh, S., Rameshan, R. and Dinesh, D.A., 2021. An Empirical Study on Machine Learning Models for Potato Leaf Disease Classification using RGB Images. In ICPRAM (pp. 515-522). 23. Oishi, Y., Habaragamuwa, H., Zhang, Y., Sugiura, R., Asano, K., Akai, K., Shibata, H. and Fujimoto, T., 2021. Automated abnormal potato plant detection system using deep learning models and portable video cameras. International Journal of Applied Earth Observation and Geoinformation, 104, p.102509. 24. Islam, M., Dinh, A., Wahid, K. and Bhowmik, P., 2017, April. Detection of potato diseases using image segmentation and multiclass support vector machine. In 2017 IEEE 30th canadian conference on electrical and computer engineering (CCECE) (pp. 1-4). IEEE. 25. N. Kaur and V. Devendran, "Ensemble Classification and Feature Extraction Based Plant Leaf Disease Recognition," 2021 9th International Conference on Reliability, Infocom Technologies and Optimization (Trends and Future Directions) (ICRITO), 2021, pp. 1-4, doi: 10.1109/ICRITO51393.2021.9596456. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 66
  • 67. 26. Hou, C., Zhuang, J., Tang, Y., He, Y., Miao, A., Huang, H. and Luo, S., 2021. Recognition of early blight and late blight diseases on potato leaves based on graph cut segmentation. Journal of Agriculture and Food Research, 5, p.100154. 27. Ashwinkumar, S., Rajagopal, S., Manimaran, V. and Jegajothi, B., 2021. Automated plant leaf disease detection and classification using optimal MobileNet based convolutional neural networks. Materials Today: Proceedings. 28. Paymode, A.S. and Malode, V.B., 2022. Transfer learning for multi-crop leaf disease image classification using convolutional neural networks VGG. Artificial Intelligence in Agriculture. 29. Rozaqi, A.J. and Sunyoto, A., 2020, November. Identification of Disease in Potato Leaves Using Convolutional Neural Network (CNN) Algorithm. In 2020 3rd International Conference on Information and Communications Technology (ICOIACT) (pp. 72- 76). IEEE. 30. Suresh A, Shunmuganathan KL (2012) Image texture classification using gray level co-occurrence matrix based statistical features. Eur J Sci Res 75:591–597 31. Tamura H, Mori S, Yamawaki T (1978) Textural Features Corresponding to Visual Perception. IEEE Trans Syst Man Cybern 8:460–473. https://doi.org/10.1109/TSMC.1978.4309999 32. Zhao M, Qiu W, Wen T, et al (2021) Feature extraction based on Gabor filter and Support Vector Machine classifier in defect analysis of Thermoelectric Cooler Component. Comput Electr Eng 92:107188. https://doi.org/10.1016/j.compeleceng.2021.107188 33. Fischetti M (2016) Fast training of Support Vector Machines with Gaussian kernel. Discret Optim 22:183–194. https://doi.org/10.1016/j.disopt.2015.03.002 34. Fan Y, Bai J, Lei X, et al (2020) Privacy preserving based logistic regression on big data. J Netw Comput Appl 171:102769. https://doi.org/10.1016/j.jnca.2020.102769 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 67
  • 68. 35. Dubey H, Pudi V (2013) Class based weighted K-Nearest neighbor over imbalance dataset. Lect Notes Comput Sci (including Subser Lect Notes Artif Intell Lect Notes Bioinformatics) 7819 LNAI:305–316. https://doi.org/10.1007/978-3-642-37456-2_26 36. Balaji VR, Suganthi ST, Rajadevi R, et al (2020) Skin disease detection and segmentation using dynamic graph cut algorithm and classification through Naive Bayes classifier. Meas J Int Meas Confed 163:107922. https://doi.org/10.1016/j.measurement.2020.107922 37. Ackora-prah J, Email T (2014) A Bilayer Feed – Forward Artificial Neural Network for Exchange Rate Prediction. Aust J Appl Math 9478:1–5 38. Singh, K., Vishwakarma, D. K., Walia, G. S., & Kapoor, R. (2016). Contrast enhancement via texture region based histogram equalization. Journal of Modern Optics, 63(15), 1444–1450. https://doi.org/10.1080/09500340.2016.1154194 39. Kumar, E. B., & Thiagarasu, V. (2018). Color channel extraction in RGB images for segmentation. Proceedings of the 2nd International Conference on Communication and Electronics Systems, ICCES 2017, 2018-Janua(Icces), 234–239. https://doi.org/10.1109/CESYS.2017.8321272 40. Michele, A., Colin, V., & Santika, D. D. (2019). Mobilenet convolutional neural networks and support vector machines for palmprint recognition. Procedia Computer Science, 157, 110–117. https://doi.org/10.1016/j.procs.2019.08.147 41. Synced. (2017). ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices. Medium. https://medium.com/syncedreview/shufflenet-an-extremely-efficient-convolutional-neural-network-for-mobile-devices- 72c6f5b01651 42. Chowdhury, N. K., Kabir, M. A., Rahman, Md. M., & Rezoana, N. (2021). ECOVNet: a highly effective ensemble based deep learning model for detecting COVID-19. PeerJ Computer Science, 7, 1–25. https://doi.org/10.7717/PEERJ-CS.551 43. Victor Ikechukwu, A., S, M., R, D., & RC, S. (2021). ResNet-50 vs VGG-19 vs Training from Scratch: A comparative analysis of the segmentation and classification of Pneumonia from chest x-ray images. Global Transitions Proceedings. https://doi.org/10.1016/j.gltp.2021.08.027 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 68
  • 69. 44. Islam, M. S., Foysal, F. A., Neehal, N., Karim, E., & Hossain, S. A. (2018). IncePTB: A CNN based classification approach for recognizing traditional Bengali games. Procedia Computer Science, 143, 595–602. https://doi.org/10.1016/j.procs.2018.10.436 45. Alex Krizhevsky, Ilya Sutskever, & Geoffrey E. Hinton. (2012). ImageNet Classification with Deep Convolutional Neural Networks. Advances in Neural Information Processing Systems, 25, 1097–1105. 46. S. Mirjalili, A. Lewis, The whale optimization algorithm, Adv. Eng. Softw. 95 (2016) 51–67. 47. Alex Lavaee, PlantifyDr Dataset | Kaggle, (n.d.). https://www.kaggle.com/lavaman151/plantifydr-dataset. 48. Sholihati RA, Sulistijono IA, Risnumawan A, et al. Potato Leaf Disease Classification Using Deep Learning Approach, In: 2020 Int. Electron. Symp., IEEE, 2020:pp. 392–397. 49. NR Deepa, N. Nagarajan, Kuan noise filter with Hough transformation based reweighted linear program boost classification for plant leaf disease detection, J. Ambient Intell. Humaniz. Comput. 12 (2021) 5979–5992. 50. M. Canayaz, MH-COVIDNet: Diagnosis of COVID-19 using deep neural networks and meta-heuristic-based feature selection on X-ray images, Biomed. Signal Process. Control. 64 (2021) 102257. 51. S. Li, H. Chen, M. Wang, A.A. Heidari, S. Mirjalili, Slime mould algorithm: A new method for stochastic optimization, Futur. Gener. Comput. Syst. 111 (2020) 300–323. 52. A. Herliana, T. Arifin, S. Susanti, AB Hikmah, Feature Selection of Diabetic Retinopathy Disease Using Particle Swarm Optimization and Neural Network, 2018 6th Int. Conf. Cyber IT Serv. Manag. CITSM 2018. (2019) 2016–2019. https://doi.org/10.1109/CITSM.2018.8674295. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 69
  • 70. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 70 1st Semester 2nd Semester 3rd Semester 4th Semester 5th Semester 6th Semester 7th Semester 8th Semester Coursework completed Comprehensive Appeared and Literature Survey started Literature survey, problem formulation Literature survey, Implementation of problem statement 1 Implementation of problem statements 1, 2 using MATLAB Implementation of problem statements 4 using MATLAB Implementation of problem statement 3 using MATLAB Synopsis and Thesis submission 9th Semester Implementation of problem statement 5 using MATLAB/Python
  • 71. 20 May 2023 D.N.KIRAN PANDIRI , Dept. of ECE, NIT Silchar 71 Thank You

Editor's Notes

  1. Agriculture is the source of food production for all the human civilization.
  2. In order to reduce the crop yield loss and usage of pesticides and fertilizers, farmers has to follow some traditional methods to take care of these which is time consuming, expensive and requires a proficient to check the farm.
  3. In order to increase the crop productivity to meet the food demand and protect the crop from diseases, farmers are using lot of pesticides and fertilizers. In the last two decades, the usage of pesticides is increased by 1/3rd which shows the impact on environment.
  4. The workforce in agriculture field is reducing day by day due to movement of the population towards urban cities. Because of the loss in production and increase in wages. People are not showing interest to do cultivation.
  5. In order to address these issues and reduce the burden on environment and farmers.
  6. By using the traditional methods, the soil type is identified based on the texture of the soil sample. Laboratory testing is time consuming, which takes 3-4 days to analyze the soil type. Hydrometer 2. Pipette 3. Sieve analysis and 1. Cone penetration 2. Vane sheer test 3. Pressure meter test 4. Dilometer test
  7. There are variety of crops being cultivated every across the globe. Among al those the main commodity crops are
  8. When comes to the research on the crops in terms of yield prediction, disease detection, crop recognition, and crop quality. These are the crops that the researchers concentrating much.
  9. For sandy soils, sieve analysis is used to identify the fraction of sand, silt, and clay, whereas for loamy and clay soils, a hydrometer test is used to determine the percentage of sand, silt, and clay.
  10. The solution acts as a dispersion agent in preventing the fine particles in suspension from clumping together.
  11. A window of dimensions 14 cm x 10 cm was made on the top of the box to place the soil sample inside the box and allow low light for capturing the images.