1. IoT BASED SOIL MONITORING SYSTEM
Monalisa Swain, Parthasarathi Nayak, Pulak Ranjan Mohanty
Regd. No.: 2001229097, 2001229103 & 2001229109
7th Semester
Under the guidance of
Shekharesh Barik
Asso. Professor
Department of Computer Science and Engineering
DRIEMS Cuttack-754022, India
1/28 Monalisa,Partha, Pulak Soil Monitoring System
2. 2/28
Outline
1. Introduction
2. Problem Statement
3. Objective
4. Related Works
5. About Dataset
6. Process Overview
7. Working of Model
8. Results
9. Conclusion
10. References
Monalisa,Partha, Pulak Soil Monitoring System
3. 3/28 Soil Monitoring System
RishubKumar
Introduction
• This system helps people for use of fertiliser according to
nuteient level.
• The major thing is watering to plant as per the requirement
with avoidance of wastage.
Soil Monitoring System
Monalisa,Partha, Pulak Soil Monitoring System
4. 4/28
Introduction
• India is said to be the diabetic capital of the world by 2030 with
over 80 million people affected by it.
• Unfortunately, more than 2/3rds of them are from the
“subaltern”.
• If identified early, this is a blindness which can be avoided.
Monalisa,Partha, Pulak Soil Monitoring System
5. 5/28
Introduction
• For rural areas (Tier 4 cities), which has small dispensaries and can
afford a fundus camera, but don’t have expert medical professionals
available.
• A tool for doctors to prioritise cases based on the severity of the
condition.
• For anyone who wants to make healthcare affordable and accurate.
Motivation
Monalisa,Partha, Pulak Soil Monitoring System
6. 6/28
Problem Statement
• India has an estimated 77 million people (1 in 11 Indians)
formally diagnosed with diabetes, which makes it the second
most affected in the world, after China. Furthermore, 500,000
Indians died of diabetes in 2020.
• To develop a tool to Grade Diabetic Retinopathy using Deep
Learning.
Monalisa,Partha, Pulak Soil Monitoring System
7. 7/28
Objective
• To identify the impacts of fertilizers and soil nutrients in
agriculture.
• Reduce the manpower and conserve water.
• Matching cases based on severity and time for immediate
diagnosis.
Monalisa,Partha, Pulak Soil Monitoring System
8. 8/28
Related Works
Sl. no. Paper publication & author Technology used Advantages Challenges
1 Published by De Gruyter Open Access
March 10, 2022
Penikalapati Pragathi and Agastyaraju
Nagaraja Rao
It combines support vector
machine, principal component
analysis, and moth-flame
optimization techniques.
Accuracy: 85.61%
outperforms other
individual algorithms.
Low accuracy.
2 Springe, 26 August 2022
Erdal Özbay
Active deep learning with CNN
(ADL-CNN) was used.
Dataset: EyePACS
Accuracy: 99.66% Not practical and slow
computation time.
3 Nor Hazlyna Harun, Zunaina Embong,
Yuhanis Yusof, Faridah Hassan
Artificial neural network (NN)
namely Multi-layered Perceptron
(MLP) trained by Levenberg
Marquardt (LM) and Bayesian
Regularization (BR) to classify the
data.
Accuracy: 72.11% (training)
and 67.47% (testing)
Very costly setup.
4 Yi-Peng Liua, Zhanqing Lib, Cong Xuc,
Jing Lid, Ronghua Lianga
Multiple weighted paths into
convolutional neural network,
called the WP-CNN was used.
Accuracy: 94.23%
weighted path
High computation time and
resource usage.
5 C. Rajaa and L. Balajib Adaptive histogram equalization
(AHE) for enhancing the input
retinal image. (CNN) and fuzzy c-
means clustering (FCM).
Accuracy: 97% More training & pre-
processing time
Monalisa,Partha, Pulak Soil Monitoring System
9. 9/28
About Dataset
• Name: DIARETDB1 - Standard Diabetic
Retinopathy Database, Size: 270MB
• Source:
• Name: Diabetic Retinopathy 224x224
(2019 Data), Size: 250MB
• Source:
The quality and quantity of the dataset has a direct effect on the decision-
making process of the ML model. And these two factors influence the
robustness, precision and performance of ML algorithms.
Dataset Used:
Monalisa,Partha, Pulak Soil Monitoring System
13. 13/28
Working of Model
• Represent input data as multidimensional arrays
image = [[[255, 255, 0], [127, 127, 127], …]]
• Predict outputs using a (parameterized) deep neural network
• Loss function depends smoothly on the parameters + tells how good our
predictions are
• Use (calculus + greediness + cleverness) to find parameters that minimize
"loss"
Deep Learning in one slide
Monalisa,Partha, Pulak Soil Monitoring System
15. 15/28
Working of Model
Image
Preprocessing
ReLU Max Pooling
CNN
Output as 1D
Vector
Sigmoid Function
on Final Output
Classification is
done for Diabetic
or Adiabatic
Retinopathy
Monalisa,Partha, Pulak Soil Monitoring System
23. 23/28
Result
• An accuracy of 98% was achieved during the experimental
testing.
• This accuracy is almost near to human (doctor) level
observation.
• The computation time on any lower end machines are very
optimal.
Monalisa,Partha, Pulak Soil Monitoring System
25. 25/28
Conclusion & Future Scope
• This deep learning model can be helpful to everyone specially rural area population where clicical
check-ups or medical machinery is uncommon.
• It can be a major step towards saving millions of life. Early stage detection and prevention will
much better in terms of precaution and treatment.
• This model is capable for producing a very satisfying and trust worthy result, in very less time.
• Deploy on edge devices, make it easily accessible for everyone.
• In future this project will be to detect the stage of diabetic retinopathy with an advanced deep
learning model. Stages like Mild, Moderate, Severe, Proliferate will be detected.
Monalisa,Partha, Pulak Soil Monitoring System
26. 26/28
References
1. Pragathi, P. and Rao, A.N., 2022. An effective integrated machine learning approach for detecting diabetic
retinopathy. Open Computer Science, 12(1), pp.83-91.
2. Dayana, A.M. and Emmanuel, W.R., 2022. Deep learning enabled optimized feature selection and
classification for grading diabetic retinopathy severity in the fundus image. Neural Computing and
Applications, pp.1-21.
3. Parthasharathi, G. U., R. Premnivas, and K. Jasmine. "Diabetic Retinopathy Detection Using Machine
Learning." Journal of Innovative Image Processing 4, no. 1 (2022): 26-33.
4. Elaouaber, Z.A., Feroui, A., Lazouni, M.E.A. and Messadi, M., 2022. Blood vessel segmentation using deep
learning architectures for aid diagnosis of diabetic retinopathy. Computer Methods in Biomechanics and
Biomedical Engineering: Imaging & Visualization, pp.1-15.
5. Dayana, A.M. and Emmanuel, W.R., 2022. An enhanced swarm optimization-based deep neural network for
diabetic retinopathy classification in fundus images. Multimedia Tools and Applications, pp.1-32.
6. Özbay, E., 2022. An active deep learning method for diabetic retinopathy detection in segmented fundus
images using artificial bee colony algorithm. Artificial Intelligence Review, pp.1-28.
7. Image Analysis & Stereology, v. 33, n. 3, p. 231-234, aug. 2014. ISSN 1854-5165.
Monalisa,Partha, Pulak Soil Monitoring System
28. 28/28
Thank You
“Any significant advancement in computer science will be
indistinguishable from magic!”
–Arthur C Clarke
Monalisa,Partha, Pulak Soil Monitoring System