Introduction
Good afternoon, everyone. We’re excited to present our paper on project titled
AgriDiagnostics – A Crop & Fertilizer Recommender System with Tomato Leaf Disease
Detection.
Literature Survey
To build a strong foundation for our project, we explored several research papers and
existing systems that focus on crop and fertilizer recommendation, as well as plant disease
detection.
The first paper, "Crop & Fertilizer Recommendation using Machine Learning" by S. M.
Melasagare et al., explores how machine learning can predict the best crops and fertilizers
using soil and weather data. The authors implemented Random Forest (RF), Support
Vector Machine (SVM), and Decision Trees, integrating IoT sensors for real-time analysis.
The second paper by Aryan Chaudhary et al. discusses crop disease detection using deep
learning models. Here, Convolutional Neural Networks achieved a 98.4% accuracy in
disease detection, while LSTM networks helped analyze disease progression with 70%
accuracy.
Problem Formulation
While reviewing the existing research and models from the literature survey, we identified
several gaps and limitations in current crop and disease prediction systems:
Limited Integration of Multiple Services: Most papers focus only on crop or fertilizer
recommendation or only on disease detection. There is no unified platform that provides all
three: crop recommendation, fertilizer suggestion, and plant disease detection in one place.
Lack of Real-Time Weather Integration: Lack of real-time weather data integration through
APIs limits practical use.
No User-Friendly Interface for Farmers: High-accuracy models exist but no user-friendly
web apps are available for farmers.
Neglected Regional Specificity: No adaptation to regional factors like state-wise climate,
soil, or crop patter
Objectives:
Agriculture is the backbone of our economy, yet farmers often struggle with key decisions
like which crops to grow, what fertilizers to use, and how to manage plant diseases
efficiently. Our project, AgriDiagnostics, aims to bridge this gap by leveraging technology to
provide data-driven recommendations in three crucial areas.
First, Crop recommendations based on soil health and environmental data,
Second, Fertilizer suggestions based on nutrient levels and the specific crop, and
And third, Tomato leaf disease detection through image processing and deep
learning.
System Architecture:
The Crop Recommendation Module helps farmers select the best crop based on soil
condition and location. The user enters values like Nitrogen, Phosphorus, Potassium, pH
level, Rainfall, State, and City. The system processes this input, checks its completeness,
and prepares it for prediction. The dataset used for training is split into training and testing
sets to build and validate the model. Important features are extracted to focus only on
useful information. Finally, the system uses the Random Forest algorithm, which creates
multiple decision trees, and the most suggested crop is given as the final result.
The Fertilizer Recommendation Module helps users identify the right fertilizer based on the
selected crop and soil nutrient values. The user enters NPK values and crop name. The
system processes this input by comparing it with ideal NPK values stored in a dataset. It
calculates the difference to find out which one is most unbalanced. The feature with the
biggest difference is selected, and the system checks whether it is too low or too high.
Using rule-based logic, a key like "Nlow" or "Phigh" is generated and matched with a
suggestion from a dictionary.
The Plant Disease Detection Module checks if a tomato leaf is healthy or infected by
analyzing a test image selected by the user. The system first pre-processes the image by
resizing and removing background noise, followed by segmentation to focus only on the
leaf area. It then extracts important features like color, shape, and texture to detect signs of
disease. Using the ResNet-9 deep learning model (a type of CNN), the image is compared
with thousands of trained examples. The system finally displays the disease name along
with a helpful treatment suggestion.
Analysis – Comparison
“Our strengths include:
High accuracy in crop and disease prediction.
Fast and accurate fertilizer recommendation.
Simple, easy-to-use web interface.
However, our system is currently limited to tomato diseases and covers only two
states.
Weather data integration is in progress and will be improved in future updates.”
Case Studies
“To understand real-world impact, we tested our system with real farmers.
In Pune, one farmer used our crop module and increased his yield.
Another farmer detected early signs of tomato disease, treated it in time, and
avoided crop loss.
These cases show how our system can make a difference on the ground.”
Conclusion
“To wrap up — AgriDiagnostics is more than just a tech project. It’s a step toward
making agriculture smarter, more sustainable, and more inclusive.
One of the biggest strengths of our system is its simplicity and accessibility
Our goal is to help farmers improve yield, reduce costs, and increase profits, while
also promoting environmentally friendly practices.
Thank you for your attention — and we hope AgriDiagnostics inspires a new way of
thinking about farming.”
Future Scope
“To take AgriDiagnostics to the next level, we have several ideas planned for future
development.
First, we aim to integrate IoT devices such as soil moisture sensors, weather stations, and
crop health monitors. Next, we plan to expand our datasets by including more crop types,
disease categories, and region-specific soil data. Another major step will be the
development of a mobile application. In short, our goal is to make AgriDiagnostics not just
a smart recommendation tool, but a complete agricultural assistant.”
