CROPSHIELD.

AI
Crop Disease Detecting AI
Presented by:

Team Member Tasks Accomplished

Jay Soni Business Report
Sumit Tabhane Product Report
Kajal Gaikwad Product Report
Adi Antulay Business Report

PROBLEM STATEMENT:

The agricultural sector, a cornerstone of global food security and economic stability, faces a
critical challenge in combating crop diseases. These diseases, exacerbated by factors such as
climate change, mono-cropping practices, and insufficient disease management systems,
significantly reduce crop yields, inflict substantial economic losses, and threaten global food
supply chains. Traditional methods for detecting crop diseases rely heavily on visual
inspections and expertise, which are often inconsistent, time-consuming, and prone to error.
This lack of precision leads to delayed interventions, inefficient resource utilization, and
heightened vulnerability for smallholder farmers who form the backbone of global agriculture.

Current solutions are either prohibitively expensive, overly complex, or poorly adapted to the
needs of farmers in resource-constrained settings. As a result, small-scale farmers, in particular,
are unable to access timely, accurate disease diagnostics or actionable insights, further
amplifying their risk of crop failure and economic instability.

To address this pressing issue, there is a need for an innovative, scalable, and cost-effective
solution that integrates artificial intelligence (AI) and computer vision to enable early and
accurate detection of crop diseases. Such a system would empower farmers with actionable
recommendations, reduce crop losses, optimize resource use, and contribute to a more
sustainable and resilient agricultural ecosystem, ultimately enhancing global food security.
PRESENT MARKET OVERVIEW:

The agricultural supply chain is plagued by inefficiencies that hinder profitability,

sustainability, and customer satisfaction. Traditional supply chains involve multiple
intermediaries, inflating costs, delaying deliveries, and reducing produce quality. Farmers lack
visibility into consumer demand, leading to overproduction or underproduction, food waste,
and lower profits.

Many farmers also lack access to advanced technologies for inventory management, demand
forecasting, and order tracking. This limits their ability to optimize operations and compete in
demand-driven markets.

Transparency is another challenge, as consumers often lack information about the source,
quality, and sustainability of their food. This undermines trust, especially as demand grows for
ethically sourced and fresh produce.

Food waste and sustainability concerns are significant, driven by poor demand forecasting,
logistical delays, and inefficient inventory management. These issues not only hurt profitability
but also increase environmental impact, including higher carbon emissions.

Finally, existing solutions often fail to offer diverse, fresh, and locally sourced options, leaving
consumers underserved. A comprehensive solution addressing these inefficiencies, enhancing
transparency, reducing waste, and expanding consumer choices is crucial to modernize the
agricultural supply chain.

PRODUCT INTRODUCTION

Our product is a cutting-edge AI-powered platform designed to address one of agriculture's

most pressing challenges: managing crop diseases effectively and sustainably. Leveraging
advanced computer vision and an inbuilt machine learning model, the system can analyze
images of crops to accurately identify diseases with remarkable precision. Once a disease is
detected, the platform provides tailored, actionable recommendations for treatment, ensuring
farmers can intervene promptly and effectively to minimize crop losses and safeguard their
yields.

But our solution doesn’t stop at disease detection. It is designed to serve as a comprehensive
agricultural companion, offering farmers detailed insights into sustainable farming practices
and advanced agricultural methods. These insights include best practices for soil health
management, crop rotation strategies, pest control techniques, and water-efficient irrigation
systems. By integrating disease diagnosis with broader agricultural knowledge, the platform
empowers farmers to make informed decisions that enhance productivity while reducing
environmental impact.

The product is user-friendly, scalable, and cost-effective, making it accessible to farmers across
varying scales of operation. By combining state-of-the-art technology with practical
agricultural solutions, it bridges the gap between traditional farming methods and modern
innovation, fostering a smarter, more resilient agricultural ecosystem. This tool is not just about
managing challenges—it’s about enabling farmers to thrive in an increasingly complex
agricultural landscape.

BUSINESS NEED ASSESSMENT

The agricultural industry faces multifaceted challenges that significantly impact productivity,
food security, and economic stability. Crop diseases remain one of the most critical threats,
causing substantial crop losses and financial hardships, particularly for smallholder farmers
who form the backbone of global agriculture. With increasing climate variability,
monocropping practices, and the lack of robust disease management systems, there is an urgent
need for innovative, scalable, and cost-effective solutions.

1. Market Dynamics
- Growing Threat of Crop Diseases: Crop diseases account for significant annual losses
worldwide, with limited access to timely and accurate detection tools exacerbating the issue.
- Demand for Precision Agriculture: As global food demand grows, farmers seek advanced
tools for efficient disease management, resource optimization, and yield improvement.
- Technology Integration Gap: Many regions, especially in developing economies, lack
access to affordable, user-friendly agricultural technologies, leaving farmers dependent on
traditional methods prone to delays and errors.

2. Key Customer Pain Points

1. Delayed and Inaccurate Disease Diagnosis
- Farmers rely on manual inspection and experience-based judgment, which often results in
misdiagnosis or delayed interventions.
- Consequences include crop losses, increased disease spread, and reduced income.

2. Lack of Actionable Insights

- Diagnosis alone is insufficient; farmers need clear, practical steps to address crop diseases,
including information on treatments, dosages, and preventative measures.

3. High Costs and Accessibility Barriers

 - Many existing solutions are expensive or require high technical expertise, rendering them
inaccessible to smallholder farmers.
- Limited connectivity in rural areas further complicates access to advanced tools.

4. Regional and Environmental Variability

- Disease prevalence and management strategies vary by region, crop type, and climatic
conditions, necessitating highly localized solutions.

3. Business Requirements
- Early and Accurate Detection
- A tool capable of diagnosing crop diseases swiftly and precisely to enable timely
interventions.
- Actionable Recommendations
- Providing farmers with step-by-step guidance on effective treatment methods tailored to
crop type, disease severity, and local conditions.
- Affordability and Accessibility
- Ensuring cost-effective solutions with features like offline functionality and multi-language
support for farmers in underserved regions.
- Scalability
- Adapting to a wide range of crops, diseases, and regional needs, ensuring the tool remains
relevant in diverse agricultural contexts.
- Sustainability
- Promoting environmentally friendly practices, reducing waste, and optimizing the use of
resources like water and fertilizers.

4. Market Opportunity
The precision agriculture market is projected to grow significantly, driven by rising demand
for sustainable farming solutions and technology integration in agriculture. By addressing
critical pain points with AI-driven crop disease detection and actionable insights, the product
has the potential to tap into a market poised for disruption.

5. Competitive Advantages
- Localized Expertise: Tailored recommendations based on regional climates, crop varieties,
and disease patterns.
- Actionable Insights: Practical steps for disease management, including organic and chemical
treatments, pest control, and irrigation adjustments.
- Data-Driven Predictions: Leveraging AI to identify emerging disease trends and provide
predictive analytics for proactive management.
- User-Centric Design: A simple, intuitive interface accessible to non-technical users, with
support for low-connectivity environments.
By addressing the urgent needs of farmers, agricultural cooperatives, and policymakers, this
solution bridges the gap between traditional methods and modern agricultural practices. Its
focus on early detection, actionable recommendations, and affordability ensures it stands out
as a transformative tool in global agriculture.

Target Audience
Audience Characteristics
1. Farmers: Smallholder & Subsistence Farmers: Need cost-effective, easy-to-use
disease management tools to safeguard their crops.
2. Large-Scale Farmers: Require scalable solutions for disease monitoring and analytics
across multiple regions.
3. Agricultural Cooperatives: Represent groups of farmers, seeking centralized tools for
disease detection, best practice dissemination, and resource optimization.
4. Agribusinesses & Enterprises: Focused on advanced analytics, integration with
existing systems, and monitoring disease trends at scale.
5. Rural & Urban Farmers: Limited access to expert advice or digital literacy; require
localized, offline-enabled, and language-friendly interfaces.
User Needs
1. Accessibility: Offline functionality, multi-device compatibility, and support for diverse
languages.
2. Customization: Tailored alerts and insights based on crops, regional conditions, and
farming practices.
3. Cost-Effectiveness: Affordable pricing models for small and large-scale operations.
Pain Points
1. Lack of expert advice for timely action.
2. High cost and complexity of current solutions.
3. Limited adaptability to local conditions and diverse crop types.
Community Features
1. Forums and collaborative tools for knowledge sharing.
2. Real-time disease alerts and updates to foster proactive engagement.
3. aims to deliver a user-centric, accessible, and scalable solution, empowering its diverse
audience to adopt sustainable and efficient agricultural practices.
EXTERNAL RESEARCH

1. Early Crop Disease Detection with AI: Strategies for Prevention

https://www.xenonstack.com/use-cases/crop-disease-detection-with-ai

2. Image‐Based Crop Disease Detection Using Machine Learning

https://bsppjournals.onlinelibrary.wiley.com/doi/10.1111/ppa.14006

3. The Era of Precision Agriculture Takes Shape

https://www.aem.org/news/the-era-of-precision-agriculture-has-arrived

4. Pathogen or Plant Disease Detection and Monitoring Market Analysis

https://bisresearch.com/industry-report/plant-disease-detection-monitoring-
market.html

5. Agtech: Breaking Down the Farmer Adoption Dilemma

https://www.mckinsey.com/industries/agriculture/our-insights/agtech-
breaking-down-the-farmer-adoption-dilemma

6. Plant Disease Image Dataset

https://www.kaggle.com/datasets/vipoooool/new-plant-diseases-dataset

These resources offer comprehensive information on the current landscape and advancements
in AI-driven crop disease detection and precision agriculture, providing a solid foundation for
external research in this field.
ML MODEL DEVELOPMENT

The CNN architecture is designed to perform multiclass classification with 38 output classes,
making it suitable for applications like image-based crop disease detection. The architecture
balances complexity and computational efficiency by utilizing multiple convolutional,
pooling, and fully connected layers. Below is a detailed explanation of each component:

1. Input Layer
Input Shape: Images of size 224x224 pixels with 3 color channels (RGB).
This defines the input dimensions for the model.

2. Convolutional Layers
Convolutional layers are the core of the CNN, responsible for feature extraction from the
input images. Each layer uses filters (kernels) to detect specific patterns such as edges,
textures, or more complex structures.

Conv1:
 Filters: 32
 Kernel Size: 7x7
 Activation Function: ReLU
 Strides: 1
 Padding: Same
 Captures initial features such as edges and corners.

Conv2:
 Filters: 64
 Kernel Size: 5x5
 Activation Function: ReLU
 Strides: 1
 Padding: Same
 Extracts intermediate-level features like textures.

Conv3:
 Filters: 128
 Kernel Size: 3x3
 Activation Function: ReLU
 Strides: 1
 Padding: Same
  Captures more abstract and detailed features without reducing spatial dimensions
since no pooling layer follows.

Conv4:
 Filters: 256
 Kernel Size: 3x3
 Activation Function: ReLU
 Strides: 1
 Padding: Same
 Extracts deeper, high-level features and prepares the data for classification.
 Pooling Layers
 Pooling layers reduce spatial dimensions to decrease computation and retain
significant features.

3. Pool1, Pool2, Pool3:

 Pooling Type: MaxPooling
 Window Size: 2x2
 Reduces the size of feature maps by half after each corresponding convolutional layer,
retaining the most significant features.

4. Flatten Layer
 Converts the 2D feature maps into a 1D feature vector.
 This vector is used as input for the fully connected layers.

5. Fully Connected Layers

These layers perform the classification task based on the extracted features.

Dense1:
 Neurons: 128
 Activation Function: ReLU
 Processes high-level abstract features.
Dense2:
 Neurons: 64
 Activation Function: ReLU
 Refines the extracted features further.
Dropout:
Prevents overfitting by randomly deactivating 50% of neurons during training.

Output Layer:
 Neurons: 38 (equal to the number of classes)
 Activation Function: Softmax
 Provides the probabilities for each class, making it suitable for multiclass
classification.

The architecture combines convolutional and pooling layers to extract hierarchical features
from input images. Fully connected layers with dropout ensure effective classification and
reduced overfitting. The final layer uses softmax for multiclass probability distribution.

CNN ARCHITECTURE
CODE:
Product Prototype
This system leverages mobile-cloud architecture and AI-driven crop disease classification to
provide an end-to-end solution for early and accurate crop disease detection. The process
integrates the ease of a mobile application with the robustness of cloud-based AI models and
actionable insights for farmers.

1. Photo Capture (Farmer Interaction)

Farmers use a mobile device to take a clear photo of a crop leaf or plant affected by potential
diseases. This step enables an easy and accessible way for farmers to initiate the disease
detection process without requiring expensive equipment or specialized knowledge.

2. Image Transmission to the Cloud

The captured image is uploaded to the cloud using the mobile app. This ensures that the
analysis is not constrained by the computational power of the farmer's device, making it
accessible for even low-end smartphones.

3. Cloud-Server Processing
A web server receives the image and manages the backend processes. This includes cleaning
the image, preparing it for analysis, and ensuring data security during the transmission.

4. Database Storage
The image is stored in a centralized database for future reference and model improvement.
This step ensures scalability, allowing the system to maintain a growing repository of crop
disease data, which can be used for training and improving AI models.

5. AI-Driven Classification
Pre-trained machine learning models analyze the image to identify crop diseases. These
models are trained on large datasets of crop images and can accurately detect diseases based
on visual patterns like discoloration, lesions, or texture changes. The system also considers
environmental factors (e.g., location, climate conditions) to improve accuracy.

6. Result Generation
The system generates disease identification results, including the name of the disease,
severity level, and affected plant parts. The system also provides actionable recommendations
for disease management, such as pesticide suggestions, organic remedies, or cultural practices
to mitigate the issue.
7. Results Sent to Farmers
The results, along with actionable insights, are sent back to the farmer's mobile device in a
simple and comprehensible format (e.g., local language support, visual guides). Farmers
receive targeted advice for timely intervention, reducing delays and inefficiencies.

BUSINESS MODEL

Monetization Strategy
CropShield employs a diversified revenue approach to ensure scalability and value for its
user base, including farmers, cooperatives, and agribusinesses:

1. Freemium Model: Offers free access to basic disease detection capabilities,

encouraging adoption.
2. Premium Services: Advanced analytics, IoT integration, real-time expert support,
customizable alerts, and comprehensive historical data, available via subscription.

Subscription Plans

1. Individual Plans: Affordable options for small farmers, focusing on productivity and
disease management.
2. Cooperative Plans: Group discounts, dashboards, and bulk licenses for farming
collectives.
3. Enterprise Plans: Tailored features for large agribusinesses, including software
integration and dedicated support.

Advertising & Sponsorship

Revenue through targeted ads and sponsored educational content related to agriculture.

The market forecast values for the gross production value of the AgTech sector are:
 YEAR Gross Production Value (Billion USD)
2025 530.81
2026 559.10
2027 582.66
2028 595.97
2029 598.69

Observations:
The market grows steadily but tapers off, indicating a potential saturation effect by 2029.
CAGR (Compound Annual Growth Rate) over this period:
598.69 1/4
CAGR = (530.81) − 1 = 3.04%
The market grows at approximately 3.04% annually.

Adjusting Financial Assumptions

1. Product Pricing and Cost Assumptions:
 Product Price: ₹500/unit
 Cost of Operation: ₹20,000/month
2. Market Penetration:
 Market size: Gross Production Value × Penetration Percentage
 For each year, we estimate penetration starting from 0.005%
(small-scale entry) and scaling up.
Revised Financial Equation
The monthly revenue equation becomes:

𝑦 = 500 𝑥 −20000
Where, 𝑥 is the total number of units sold per month. The value of x depends on market
penetration.

Revenue Projection: Year-wise Estimates

Year Market Value(Billion Penetration(%) Units Sold Revenue(Million

USD) (Million) Rupees)
2025 530.81 0.005% 0.02654 12.27
2026 559.10 0.01% 0.05591 27.96
2027 582.66 0.02% 0.11653 58.26
2028 595.97 0.03% 0.17879 89.4
2029 598.69 0.05% 0.29935 148.68
Example Calculations:

For 2025:
Units Sold = 530.81B×0.005%=0.02654M (approx. 26,540 units).
Revenue = 500×26540−20000=₹12,27,000.

For 2029:
Units Sold = 598.69B×0.05%=0.29935M (approx. 299,350 units).
Revenue = 500×299350−20000=₹1,48,68,000.

Insights and Recommendations

 Growth Trends: While the market is growing, the CAGR indicates a slowing pace
from 2028 onward. Early penetration and scaling strategies are crucial to capture
market share before saturation.
 Break-even Analysis: With 40 units/month, the business achieves break-even at a
₹500/unit price point:
500×40−20000=0
 Opportunities: Partnering with government initiatives and leveraging digital
platforms can improve penetration rates.