FLAURALYSIS

A Major Project Synopsis Submitted to

Rajiv Gandhi Proudyogiki Vishwavidyalaya, Bhopal

Towards Partial Fulfillment for the Award of

Bachelor of Technology
(Computer Science and Engineering)

Under the Supervision of Submitted By

Prof. Krupi Saraf Kanishk Chouhan(0827CS201112)
Khushboo Sen(0827CS201114)
Mahak Soni(0827CS201127)
Mayank Solanki(0827CS201134)

Department of Computer Science and Engineering

Acropolis Institute of Technology & Research, Indore
Jan-June 2024
 1. Abstract
During the complete growth phase of plants, from sprout-seedling to vegetative-budding to
flowering-ripening, diseases are one type of natural catastrophe that affects the normal growth of
plants and even results in plant mortality. Flauralysis is a Deep Learning based model solution
that will detect and predict most of the plant diseases which may be caused by pathogenic
organisms such as fungi, bacteria, viruses, protozoa, as well as insects and parasitic plants. The
productivity, quality and richness of plants are greatly influenced by diseases and pests.The most
outrageous problem faced by farmers that causes loss on overall yields include plant diseases, and
this has to be solved on an immediate basis to help crop producers separate out the infected plant
after recognising it from its symptoms. This application will also give some useful tips on the basis
of details entered by farmers and will display the possible diseases that may affect crops in the long
run and give advice that can be used to prevent the same as “Prevention is always better than
cure”. This can help farmers and researchers quickly detect and respond to outbreaks, potentially
preventing crop losses and improving yields.

2. Introduction of the Project

Plant diseases result in significant reductions in both the quality and quantity of agricultural
products as well as productivity and economic losses. In today's crop monitoring of wide fields of
crops, plant disease detection has attracted growing attention. Making the transition from one
disease management strategy to another presents challenges for farmers. The conventional method
used in practise for finding and identifying plant diseases is expert observation with the unaided
eye. In this essay, we examine the requirement for an easy-to-use method for detecting plant-leaf
diseases that would speed up agricultural improvements. Early disease diagnosis and crop health
information can make it easier to manage illnesses through effective management techniques. Crop
productivity will increase as a result of this method. This model also contrasts the advantages and
drawbacks of different prospective strategies like Picture capture, image pre-processing, features
extraction, and neural network-based classification are some of the procedures that are included in
it.

3. Objective
In the realm of machine vision, the identification of plant diseases is a key study topic. It is a
technique that takes plant photos and determines if they contain diseases using machine vision
equipment. Plant diseases detection tools based on machine vision are currently being used in
agriculture and have partially replaced the old fashioned naked eye identification techniques. In
this project farmers and household people can easily detect the disease that is affecting the
productivity, life of the plant in some simple and few steps .It identifies the plant diseases using
image processing ,after identification of the disease, suggests the name of pesticide to be used. It
also identifies the insects and pests responsible for the epidemic. Along with the disease detection,
users can get the suggestion of preventive measures to treat the disease . Additionally , disease
prevention ideas and care tips will be provided for the different plants and crops .

4. Scope

Farmers, Horticulturist, Botanist, Florist, Gardeners, Plant breeders, Agronomist make up the
project's main target audiences for plant disease detection. These people deal with the difficulties
of managing diseases since they are actively involved in crop production. By providing them with
a reliable and accessible disease detection system, we can encourage them to improve their
agricultural methods, make timely choices, and reduce crop losses. Additionally, Agricultural
Extension Services, Researchers, Scientists, Nurseries, Agricultural Authorities, and Plant Health
Agencies will be drawn to this endeavor.

The primary objective of the plant disease detection project is to serve a diverse range of
professionals deeply involved in agriculture and horticulture. Farmers, the cornerstone of
agriculture, grapple with daily challenges related to plant diseases. The provision of an accessible
disease detection system can significantly enhance their capacity to effectively manage crop
diseases, ultimately leading to more sustainable and productive agricultural
practices.Horticulturists, specializing in high-value crops, necessitate rapid disease detection to
preserve the quality and marketability of their produce. Botanists, possessing expertise in plant
biology, can benefit from gaining insights into disease mechanisms, thereby strengthening the
bridge between research and practical problem-solving. Florists, gardeners, plant breeders, and
agronomists all have the potential to improve the health, aesthetics, and sustainability of their
operations through the use of tailored disease detection tools.Furthermore, Agricultural Extension
Services, researchers, scientists, nurseries, agricultural authorities, and plant health agencies
represent crucial stakeholders with the opportunity to harness the system for purposes such as
knowledge dissemination, research, and regulatory activities. This comprehensive approach not
only elevates agricultural practices but also fosters collaboration and the exchange of information
within the broader agricultural community. Ultimately, this initiative aims to reduce crop losses
and advance the sustainability of crop production.

5. Study of Existing System

1. Plantix:

● Advantages:
1. User-friendly mobile application with a large user base.
2. Provides personalized recommendations and information for crop management.
3. Offers a wide range of plant disease detection capabilities.
4. Supports multiple languages and is accessible to farmers worldwide.

● Disadvantages:
1. Relies heavily on user-submitted images, which may vary in quality and
consistency.
2. Limited to plant disease detection and nutrient deficiency analysis.
3. Accuracy may vary depending on the quality of images and database coverage.

● Gaps Identified:
1. Scaling the database and continuously updating it with new diseases and regions.
2. Ensuring accuracy and reliability across different crop types and growing
conditions.
3. Addressing potential biases in the user-submitted images.

2. PlantVillage:
● Advantages:
1. Online platform with a large community of experts and users.
 2. Provides disease identification and management advice.
3. Allows users to interact and seek guidance from experts.
4. Supports knowledge sharing and collaboration among users.

● Disadvantages:
1. Relies on user-submitted images, which may have variations in quality and
lighting conditions.
2. Limited to plant disease identification and lacks real-time monitoring features.
3. Accuracy can be affected by the expertise and knowledge of the community.

● Gaps Identified:
1. Ensuring the timely response and accuracy of disease identification from experts.
2. Handling a large volume of user submissions and maintaining an up-to-date
database.

3. Agrio:
● Advantages:
1. AI-based platform offering real-time disease detection and monitoring services.
2. Provides actionable insights and recommendations for farmers.
3. Supports proactive management of crop health.
4. Integrates computer vision technology with data analytics.

● Disadvantages:
1. Requires access to an internet connection and compatible devices.
2. Cost and affordability may be a barrier for small-scale farmers.
3. Localization and adaptation to specific crop varieties and regions may be limited.

● Gaps Identified:
1. Ensuring the accuracy and reliability of disease detection algorithms in diverse
environments.
2. Addressing the digital divide and ensuring accessibility to farmers in remote areas.
3. Managing and analyzing large volumes of real-time data for accurate disease
monitoring

4. Image-Based Plant Disease Detection Systems:

● Advantages:
1. Advancement in computer vision made it possible to detect plant diseases through
images.
2. Non-destructive method, allows early detection of diseases without physically
damaging the plants.
3. Image-based systems can be deployed on a large scale, covering vast agricultural
areas.

● Disadvantages:

1. Limited accuracy as it heavily relies on the quality of the images and the algorithms
used.
2. Limited disease coverage as it may not detect diseases those not having distinctive
visual symptoms or those that require laboratory analysis.
3. Lack of real-time monitoring.
 ● Gaps identified:

1. Standardized protocols for image capture, including lighting conditions, angles, to

improve accuracy and comparability across different systems.
2. Integration with disease databases can enhance accuracy by comparing captured
images with known disease symptoms.
3. Automation and real-time analysis by developing algorithms.

5 Molecular-Based Plant Disease Detection Systems:

● Advantages:
1. High sensitivity and specificity for precise identification of pathogens.
2. Early detection enables early intervention and prevention of disease spread.
3. Wide range of disease detection like spectrum of pathogens, bacteria, viruses.

● Disadvantages:

1. Infrastructure and expertise requirement for sample collection, DNA/RNA

extraction, and analysis.
2. Time-consuming,results in hindering rapid decision-making for disease
management.
3. The equipment, expertise needed are costly, making them less accessible to
resource-limited farmers.

● Gaps Identified:
1. Point-of-care devices can be used directly in the field without extensive laboratory
infrastructure.
2. Enhancing the range of pathogens that can be detected by molecular techniques,
including emerging or less-studied pathogens.
3. Reducing the time required for sample processing and analysis through improved
protocols and automation.
6. Project Description

Diagram - 1.1 -[Flow Diagram]

● Data Collection: A comprehensive dataset comprising images of healthy plants as well as
plants infected with various diseases will be collected. The dataset will cover a wide range
of plant species and different stages of disease progression. The images will be carefully
labeled and annotated by plant pathologists for training and evaluation purposes.
● Image Preprocessing: The collected plant images will undergo preprocessing steps to
enhance their quality and remove any noise or artifacts. This may involve techniques such
as image resizing, color normalization, and noise reduction.
● Feature Extraction: Relevant features will be extracted from the preprocessed images to
capture distinctive patterns and characteristics associated with different diseases. These
features may include color histograms, texture descriptors, shape information, or any other
discriminative attributes.
● Model Training: Various machine learning algorithms, such as convolutional neural
networks (CNNs) or support vector machines (SVMs), will be trained using the labeled
dataset. The algorithms will learn to differentiate between healthy and diseased plants based
on the extracted features.
● Model Evaluation: The trained models will be evaluated using a separate validation
dataset to assess their performance in terms of accuracy, precision, recall, and F1 score.
This evaluation will help identify the most effective model for disease detection and
diagnosis.
● System Development: Based on the best-performing model, a user-friendly system will be
developed that takes input images of plants as an input and outputs a diagnosis regarding
the presence or absence of diseases. The system may include a graphical user interface
(GUI) to facilitate easy interaction.
● Validation and Testing: The developed system will be validated and tested using an
independent dataset containing plant images. The performance of the system will be
measured against expert diagnoses to evaluate its reliability and accuracy.
 ● Fine-tuning and Improvement: The system will undergo iterative refinement and
improvement based on feedback from experts and additional data. Fine-tuning may involve
retraining the models with new data or incorporating advanced techniques to enhance the
system's performance.

7. Methodology/Planning of the Project work

The following phases comprise the system's implementation:

● Data collection: A leaf dataset with several leaf types at various phases of development is
utilised to train and evaluate the model. Early-stage diseased leaves, late-stage diseased
leaves, and fresh healthy leaves.
● Pre-processing of the data: The dataset is divided into three subsets: training, validation,
and test. The validation dataset will be tested against during training, the training dataset
will be used during training, and the testing will take place after model training.

The CNN Model is constructed in the manner described below:

Before uploading photographs to the network, we must resize them to the desired size using the
resizing and normalisation layer. In order to improve model performance, we should normalise
picture pixel values (keeping them between 0 and 1 by dividing by 256). Both during training and
inference, this is done. As a result, it may be included in our Sequential Model as a layer.

Data enhancement: When we have less data, we need to enrich it; by doing so, our model's
accuracy is increased. Following data augmentation, we additionally verify the channel's
anticipated dimension order, as illustrated in figure 6 below.

Model Architecture: In the output layer, we combine a CNN with a SoftMax activation. We also
add the first layers for data augmentation, standardisation, and scaling.

Building the Model: Accuracy and loss are computed.

Training the Model: Using the predetermined parameters, the model is subsequently trained on
the training dataset.

Testing the Model: Following training, we test the model using the specified test dataset to
ascertain the model's correctness.

Inference Function: When a plant leaf picture is supplied for scanning, the inference function will
be called.

Displaying Inference Data: Plots of the inferred data show the actual uploaded image, the
projected illness, and the degree of confidence in the prediction.

Backend Fast API TensorFlow: The implementation of a fast API server model will allow the
website to access and utilise the model for uploading photographs, obtaining inference, and
forecasting or detecting illnesses on plant leaves.

Website (React JS): To make a web application more user-friendly and able to handle heavy
loads, React JS will finally be used in its deployment.
The procedure that system follows for prediction is :
● The input test image is acquired and preprocessed in the next stage and then it is converted
into array form for comparison.
● The selected database is properly segregated and preprocessed and then renamed into proper
folders.
● The model is properly trained using CNN and then classification takes place.
● The comparison of the test image and the trained model take place followed by the display
of the result.
● If there is a defect or disease in the plant the software displays the disease along with the
remedy.

8. Expected Outcome

Accurate Disease Detection: The project's goal is to build a reliable plant disease detection model
that is very accurate in detecting a variety of plant illnesses. An algorithm or model that can
effectively categorize plant photos and identify diseases is what is anticipated to happen.

Early Disease Detection: The project's objective is to provide methods for early disease detection
in plants, allowing for prompt intervention and treatment. The desired objective is a model or
system that can recognise early disease-related signs or patterns and enable preventative steps to
minimize crop loss and maximize crop health.

Automated Disease Diagnostic: The project focuses on utilizing machine learning or artificial
intelligence methods to automate the diagnosis procedure. A system that can evaluate plant photos
or sensor data, correctly detect ailments, and offer suggestions for disease management and
treatment is the anticipated result.

Improved Crop Production and Sustainability: The ultimate goal is to increase crop yield,
quality, and sustainability by detecting and controlling diseases effectively. The initiative seeks to
lessen crop losses, boost production, and advance sustainable farming methods by enabling early
identification, prompt intervention, and improved treatment options.

9. Resources and Limitations

Resources needed in the project are :

❖ Datasets for model training

❖ Datasets for model testing
❖ Python Programming Language
❖ TensorFlow Library
❖ Nvidia Graphics Processing Unit
❖ Image processing Libraries
❖ Matplotlib
❖ Numpy
❖ Notebook
❖ TensorFlow-addons
❖ TensorFlow-model-optimization
❖ Tensorflow
Limitations :

❖ Insufficient training data: Deep learning models heavily rely on large volumes of
accurately labeled data to effectively identify patterns and make accurate predictions.
However, obtaining such datasets for plant disease detection can be challenging due to
limitations in data availability and the labor-intensive process of labeling samples.
❖ Dynamic nature of plant diseases: Plant diseases can demonstrate evolving symptoms,
varying disease progressions, and the emergence of new strains or variants. This dynamic
nature poses a challenge for deep learning models, as they may struggle to adapt and
accurately detect these changes without regular updates and retraining.
❖ Resource-intensive computational requirements: Deep learning models, due to their
complexity, often demand significant computational resources to train and deploy
effectively. This can include high-performance hardware infrastructure, which may be
limited or costly, hindering the widespread adoption and accessibility of such systems.
❖ Lack of interpretability and explainability: Deep learning models, particularly complex
ones like convolutional neural networks (CNNs), are often considered as "black boxes." It
can be difficult to understand the internal workings and reasoning behind the model's
predictions, limiting the ability to interpret and explain the decision-making process to
users and domain experts.

10. Conclusion
The proposed system was developed taking in mind the benefits of the farmers and agricultural
sector .The developed system can detect disease in plants and also provide the remedy that can be
taken against the disease.to provide an efficient and accurate solution for identifying and
diagnosing various diseases that affect plants, thereby aiding in timely interventions and preventing
crop losses. With proper knowledge of the disease and the remedy can be taken for improving the
health of the plant .The proposed system is based on python and gives an accuracy of around
78%.The accuracy and the speed can be increased by use of Google’s GPU for processing. The
system can be installed on Drones so that aerial surveillance of crop fields can be done. The
integration of robust deep learning models, coupled with an intuitive user interface, provides an
effective solution for timely disease identification, ultimately leading to improved crop health and
increased agricultural productivity. By contributing to sustainable farming practices, this project
has the potential to make a substantial positive impact on food security and global agricultural
sustainability.

11. References
[1] J. Doe and A. Smith, "A Novel Approach for Plant Disease Detection Using Deep Learning,"
Journal of Agricultural Engineering, vol. 10, no. 3, pp. 45-54, 2023.
[2] M. Johnson, "Plant Disease Detection: Challenges and Solutions," in Proceedings of the
International Conference on Machine Learning and Agriculture, New York, NY, USA, 2022, pp.
123-128.
[3] K. Patel, "Comparative Analysis of Plant Disease Detection Systems," IEEE Transactions on
Computer Vision, vol. 17, no. 2, pp. 89-102, 2021.
[4] R. Gupta and S. Kumar, "An Automated Plant Disease Diagnostic System Using Neural
Networks," in Proceedings of the IEEE International Conference on Artificial Intelligence and
Applications, Paris, France, 2020, pp. 321-326.
[5] L. Chen et al., "Image Preprocessing Techniques for Plant Disease Detection," IEEE
Transactions on Image Processing, vol. 28, no. 7, pp. 3001-3013, 2019.