MangoLeafBD · EfficientNetB0

AI Crop Health Decision Support

FarmAI

AI-powered crop health prediction and decision support system.

Helping farmers make data-driven decisions using machine learning.

Upload a photo of a mango leaf and the model classifies it across 8 disease categories in the browser, giving fast visual diagnosis support without a backend.

01

Upload a leaf image

Drag and drop or select a JPG/PNG photo of the mango leaf you want to analyse.

02

Instant classification

EfficientNetB0 runs client-side in your browser — no server, no upload delays.

03

Explore the results

See the prediction, confidence, disease details, and visual analysis of the leaf.

Problem → Solution → Impact

Why this system matters

Problem

Farmers face uncertainty from disease pressure, environmental shifts, and delayed field inspection, which can lead to weak planning and avoidable crop loss.

Solution

A machine learning image classifier analyzes mango leaf symptoms and predicts the most likely disease class from the uploaded image.

Impact

  • Better planning
  • Reduced risk
  • Improved efficiency

Use Cases

Where it helps

Crop health screening

Quickly check whether a mango leaf shows signs of a known disease class.

Risk analysis

Use early image-based classification as an input to field monitoring and disease management decisions.

Treatment planning

Support faster escalation from observation to action by narrowing the likely disease type.

Model Intelligence

What powers the prediction

Model used

EfficientNetB0 convolutional neural network exported to TensorFlow.js for browser inference.

Features learned

Leaf color distribution, lesion edges, surface texture, shape irregularities, and disease pattern contrast.

Dataset source

Mango leaf disease image dataset with 8 classes, prepared locally in this project for retraining and export.

Metrics

Engineering snapshot

4,000 leaf images in the local dataset
8 supported disease classes
224×224 image resolution used for inference
85–95% target accuracy range after retraining, depending on split and checkpoint

Architecture

System flow

User Input
→
Data Processing
→
ML Model
→
Prediction
→
UI

Drop an image here

or click to browse  ·  JPG, PNG supported

No image selected

Loading browser model…
Uploaded leaf

Detected Condition

—

Confidence —
Confidence is low — the model couldn't classify this image reliably. Try uploading a clearer, well-lit photo of the leaf.

Disease Details

Overview

Loading…

Symptoms

Loading…

Control

Loading…

Visual Analysis

Sobel edge detection highlights structural boundaries — veins, lesion edges, and texture transitions that the model uses as discriminating features.

RGB channel histogram. Healthy leaves skew heavily green; diseased leaves often show elevated red or reduced green intensity.