1. Nadimpalli Satyanarayana Raju Institute of Technology (NSRIT)
PLANT AILMENT DETECTION USING
DEEP LEARNING AND CONVENTIONAL NEURAL NETWORK
Team members:-
P.V.SURYA KAMAL[20nu1a4220]
NCH. DEDEEPYA [20nu1a4216]
G.JASWANTH CHOWDARY [20nu1a4206]
P.S.RATNA KUMARI [20nu1a4218]
P.SUVARNA[20nu1a4221]
FACULTY TRAINER :-
APARAJINI MAM
BRANCH:-
CSM
YEAR:-
3RD
2. Nadimpalli Satyanarayana Raju Institute of Technology (NSRIT)
Deep learning and Convolutional Neural Networks have been used to detect disease categories in unstructured data.
These models are relatively easy to train with a few hundred thousand examples, but the results may not be reliable
due to noise in images and other factors. In this paper, we consider combining both deep and shallow learning
methods using convolutional neural networks (CNN). In this work, I introduce a tree based deep learning architecture
for plant disease detection via Image and video sequences. The proposed model is trained with the Genetic
Programming (GP) algorithm to explore a well-defined optimization space and then trained minimally on the abstract
functions of the learned loss function to obtain a given number of trees. The final output of each tree is a decision in
between one positive and one negative class.Plant disease detection using deep learning and C++ CNNs
(Consequence Normal Transformation Networks) is an efficient method for detecting various diseases on crops. The
current state-of-the-art method detects the disease in crops by manual checking of concern symptoms, which takes a
lot of time and resources. The proposed method determines the disease prevalence on plants by analyzing different
types of signals and training it to classify diseases successfully.
ABSTRACT:-
4. Nadimpalli Satyanarayana Raju Institute of Technology (NSRIT)
INTRODUCTION:-
Plant diseases can occur, which is bad for agricultural production. Food insecurity will worsen if plant diseases are
not promptly discovered . Plant diseases must be prevented and controlled effectively on the basis of early detection,
and they are a critical component of agricultural production management and decision-making. Identification of plant
ailments has become a major concern in recent years. Infected plants typically exhibit glaring stains or lesions on
their leaves, stems, flowers, or fruits. Each disease or pest condition typically exhibits a distinct visible pattern that
can be used to specifically identify abnormalities. The leaves of plants are typically the main source for identifying
plant diseases, and the majority of disease symptoms may start to show on the leaves.
On-site identification of diseases and pests of fruit trees is typically done by agricultural and forestry experts, or by
farmers using their own knowledge. This approach is subjective as well as time-consuming, exhausting, and
ineffective. Very poor performance when employed alone, while efforts have been made to increase performance
through the synthesis of other techniques. entails the use of segmentation approaches, which requires the
separation of plants from their roots in order to extract geometric and related properties. applied using datasets that
have photographs that are challenging to find in the actual world.
5. Nadimpalli Satyanarayana Raju Institute of Technology (NSRIT)
Without visualization techiniques:-
In order to classify illnesses in plants and illustrate the model's importance, CNN was employed in conjunction with
histogram approaches. To diagnose tomato leaf illnesses, simple CNN architectures including AlexNet, GoogLeNet,
and ResNet were built. Plots of training and validation accuracy were used to display the model's performance;
ResNet was judged to be the best CNN architecture. LeNet architecture was utilised to detect illnesses in banana
leaves, while CA and F1-score were used to assess the model in both colour and grayscale modes. AlexNet,
AlexNetOWTbn, GoogLeNet, Overfeat, and VGG architectures were utilised among the five CNN models, with
VGG outperforming them all. In, three classifiers—Support Vector Machines (SVM), Extreme Learning Machine
(ELM), and K-Nearest Neighbor (KNN)—were coupled with cutting-edge DL models, including GoogLeNet, ResNet-
50, ResNet-101, Inception-v3, InceptionResNetv2, and SqueezeNet, to recognise eight different plant diseases.
6. Nadimpalli Satyanarayana Raju Institute of Technology (NSRIT)
With visualization techiniques:-
The subsequent methods made use of DL models/architectures as well as visualisation tools that
were developed to help people understand plant diseases better. For instance, the CaffeNet CNN
architecture helped identify 13 distinct types of plant diseases and produced a CA of 96.30%, which
was better than the previous method, SVM, for visualising the symptoms of plant sickness. The illness
spots were also indicated using a number of filters.
7. Nadimpalli Satyanarayana Raju Institute of Technology (NSRIT)
K-NEAREST NEIGHBOR (KNN) :-
An approach to supervised learning is used by the KNN classifier. In disciplines including machine learning, image
processing, and statistical estimation, this technique is frequently employed. When fresh learning data enters, this
algorithm creates a categorization of the existing learning data. The basic idea behind this approach is to place fresh
data into an existing sample set in the closest cluster. Several distance functions are used to calculate the separation
between these two data points. Euclidean distance, Minkowski distance, and Manhattan distance are the three most
well-known functions.
SUPPORT VECTOR MACHINE (SVM):-
Statistical learning theory is the foundation of the Vapnik-developed SVM technique. A linear discriminant function with
the biggest marginal separating the classes from one another is the goal of the SVM method. Support vectors refer to
the learning data that is most closely related to the hyperplane. Both linearly discernible and indistinguishable data
sets can be classified by SVM. This classifier has been effectively used to address issues in numerous fields,
including image and object identification, voice recognition, fingerprint recognition, and handwriting recognition.
8. Nadimpalli Satyanarayana Raju Institute of Technology (NSRIT)
Dataset:- plant/village
The plant /village dataset consist of 54303 healthy and unhealthy leaf images divided in to 38 categories by species
and diseases.
9. Nadimpalli Satyanarayana Raju Institute of Technology (NSRIT)
Related works:-
[1]: in this article we use a deep learning technique to identify the plant disease detection and their symptoms this helps to
identify the disease and how to solve using some techniques using CNN algorithm.
[2]: the CNN algorithm is using image from plants village dataset analysis the plant disease and their symptoms
[3]:the svm is used in forest and logistic regression have been applied . This svm is calssifed to identify the different leaf
diseases.
[4]:the knn algorithm is to find for some specific plants like(Rice leaf) this knn is finding a maximum k value.
10. Nadimpalli Satyanarayana Raju Institute of Technology (NSRIT)
Reference list:-
[1]:Muhammad Hammad Saleem, Johan Potgieter and Khalid Mahmood Arif, Plant Disease Detection and
Classification by Deep Learning, 2019, http://www.mdpi.com/journal/plants
[2]:MUAMMER TÜRKOĞLU DAVUT HANBAY, Plant disease and pest detection using deep learning-based features,
2019, https://doi.org/10.3906/elk-1809-181
[3]:Lili li, shujuan zhang, and bin wang, plant disease detection and classification by deep learning- a review,2021