The output of the study is an application called AgroAI: Leaf-Based Crop Disease Detection System. This

web-based and mobile-enabled system is designed to assist farmers in detecting crop diseases early by

providing a platform where users can upload images of crop leaves such as banana, corn, mango, potato, and

tomato. The system automatically analyzes the images using a Convolutional Neural Network (CNN) model to

identify possible diseases, provide treatment suggestions, and generate confidence scores. Farmers can also

access their diagnostic history, view data analytics on crop health, and participate in community forums for

agricultural knowledge sharing. For administrators and researchers, the system stores and organizes

responsive layout, and MySQL/MariaDB as the database management system. The frontend integrates Chart.js

for visualizing disease frequency, confidence scores, and trend data. The mobile interface uses frameworks

compatible with Android/iOS to ensure accessibility for rural farmers. To guarantee quality and effectiveness,

the system is evaluated using the ISO/IEC 25010 software quality model, which measures functionality,

usability, reliability, performance efficiency, and maintainability. This ensures that AgroAI is not only accurate

and user-friendly but also reliable, efficient, and practical for addressing the real-world needs of farmers and

agricultural communities.

The study followed an applied research approach using the Iterative System Development Life Cycle (SDLC)

model. The system is evaluated using the ISO/IEC 25010 Software Quality Model focusing on functionality,

usability, reliability, performance efficiency, and maintainability. Fifty (50) respondents participated in the

Computer Vision; Data Analytics; Mobile Agriculture Application; Web-Based System; Precision Farming;

ISO/IEC 25010; Farmers’ Decision Support System.

Accurate and timely detection of plant diseases remain a persistent challenge in global agriculture. Traditional

diagnostic approaches, which rely heavily on visual inspection by human experts, are often labor-intensive,

subjective, and prone to error (Barbedo, 2019). These limitations frequently result in delayed interventions and

increased crop losses that could otherwise be prevented with earlier diagnosis (Ferentinos, 2018). As a result,

there is growing demand for automated diagnostic tools that can deliver rapid, consistent, and scalable

assessments across diverse agricultural contexts.

Recent developments in computer vision, particularly the application of Convolutional Neural Networks

Accurate crop disease diagnosis is crucial for minimizing crop losses, yet traditional visual inspection methods

are often slow, inconsistent, and reliant on expert availability (Barbedo, 2019; Ferentinos, 2018). With

advancements in artificial intelligence (AI) and image-based diagnostics, Convolutional Neural Networks

(CNNs) have demonstrated superior performance in identifying plant diseases from leaf images—often

surpassing human accuracy (Mohanty, Hughes, & Salathé, 2016; Brahimi, Boukhalfa, & Moussaoui, 2017; Liu

et al., 2020).

Recent advancements in data analytics and machine learning have significantly improved the accuracy and

scalability of plant disease diagnosis systems. Convolutional Neural Networks, in particular, have been widely

adopted due to their strong capability in extracting hierarchical features from leaf images (Ferentinos, 2018).

CNN-based architectures such as AlexNet, VGGNet, and ResNet have been successfully used to classify plant

In areas such as Brgy. Lidong, Polangui, Albay, local farmers rely heavily on crops like banana, corn, mango,

potato, and wheat for both livelihood and food security. However, one of the major challenges faced by these

farmers is the early detection and accurate diagnosis of crop diseases. The region's tropical climate—

characterized by heavy rainfall, flooding during wet seasons, and drought during El Niño events—further

complicates plant health management and agricultural productivity.

Traditionally, farmers depend on manual inspection or informal advice for diagnosing plant issues. These

methods are often subjective, delayed, or inaccurate, leading to reduced yields, increased use of pesticides, and

economic loss. As the agricultural landscape becomes increasingly vulnerable to climate and pest stressors, the

need for intelligent, technology-driven solutions becomes more urgent.

This study proposes the development of AgroAI, a web-based application powered by Convolutional Neural

This study shows how building a system like AgroAI can help farmers easily detect plant diseases by

uploading images of crop leaves. It aims to give accurate results using Artificial Intelligence, specifically

Convolutional Neural Networks (CNNs), without needing expert knowledge. This makes the process faster

and more reliable, helping farmers take early action to avoid crop damage and reduce losses. Specific

objectives are:

through accurate diagnosis using Convolutional Neural Networks (CNNs), Computer Vision, and Data

Analytics, enhancing crop management and agricultural decision-making.

view diagnoses, and access actionable insights.

and recommendations based on user roles.

models.

patterns.

outbreaks.

training dataset.

varieties.

agronomist advice.

The AgroAI system's technical design, process optimization, and overall goals are all guided by important

passing images through layers of convolution and pooling, progressing from detecting simple features

like edges and textures to complex, disease-specific characteristics such as leaf lesions and discoloration

(LeCun et al., 2015). This approach removes the need for manual feature engineering, enabling scalable

and efficient analysis of agricultural images.

with disease categories, especially when trained on diverse datasets that include multiple crop species,

debris, which require preprocessing techniques like normalization and data augmentation to improve

on specific agricultural datasets—has proven effective in addressing limited data availability while

maintaining computational efficiency (Too et al., 2019).

This study is anchored on the Input–Process–Output (IPO) Model, which provides a systematic representation

of how the AgroAI system transforms raw data and user inputs into meaningful diagnostic results and

decision-support outputs. The model illustrates the logical flow of the system’s operations—from user

engagement to computational processing—culminating in outputs that assist farmers in managing crop health

effectively.

Figure 1: Input-Process-Output Model

The output stage provides the user with a detailed diagnosis, including the disease name, symptoms, and

confidence score. Users receive treatment recommendations and can access a personalized dashboard that

stores their diagnostic history and saved treatments. Additionally, they can view and contribute to community

discussions and access guides for disease prevention and treatment. A feedback loop is also implemented,

and user-centered.

potential of AI in addressing challenges faced by smallholder farmers, particularly in regions where vegetables

are a key source of income. By providing early detection of pests and diseases, the system helps farmers

reduce crop losses and improve yields.

According to Garcia & Reyes (2020), CNNs have shown promise in detecting diseases in fruit crops such as

mangoes and bananas. The study explored the development of an automated system for disease identification

using CNN-based models, which were trained on image datasets collected from local farms. The findings

suggest that CNNs could enhance the ability to detect diseases early, which is crucial for preventing