Skip to content

README.md

Latest commit

 

History

History
182 lines (127 loc) · 5.04 KB

README.md

File metadata and controls

182 lines (127 loc) · 5.04 KB
Raw

VegHealth

A Deep Learning project for detecting and classifying vegetable health from images.

VegHealth explores computer vision techniques for identifying whether vegetables are healthy or affected by common diseases. The project is being developed incrementally, with individual vegetable categories trained and evaluated as separate components within one larger system.

Project Overview

Plant diseases can significantly affect crop quality and agricultural productivity. Early identification can help farmers respond before diseases spread extensively.

VegHealth uses image classification and Deep Learning to investigate how computer vision can assist with vegetable disease detection.

The project currently focuses on building and evaluating models for individual vegetables before integrating them into a broader vegetable health classification system.

Current Vegetables

Vegetable Status
🍅 Tomato In development
🥔 Potato Planned
🌶️ Pepper Planned
🥬 More vegetables Planned

The project structure is designed so additional vegetables can be added without creating separate repositories.

Repository Structure

VegHealth/
│
├── Tomato/
│   ├── data/
│   ├── model.ipynb
│   ├── train.py
│   └── ...
│
├── Potato/
│   └── ...
│
├── Pepper/
│   └── ...
│
├── .gitignore
├── requirements.txt
└── README.md

Each vegetable has its own directory containing the data-processing, experimentation, training, and evaluation work associated with that vegetable.

Tomato Health Classification

The first component of VegHealth focuses on tomato health classification.

The model is trained to distinguish between different tomato health conditions using image data.

Current workflow

Image Dataset
      ↓
Data Exploration
      ↓
Preprocessing
      ↓
Train / Validation Split
      ↓
Deep Learning Model
      ↓
Training
      ↓
Evaluation
      ↓
Prediction

The tomato model is being developed and evaluated before moving on to additional vegetables.

Technologies

  • Python
  • NumPy
  • Pandas
  • Matplotlib
  • TensorFlow / Keras
  • Jupyter Notebook
  • Git & GitHub

Additional libraries will be added as the project develops.

Dataset

The project uses image datasets containing vegetable images organized according to their health condition.

Because image datasets and trained neural-network models can be very large, raw datasets and large model files are not committed directly to this repository.

The repository instead contains the code and notebooks required to work with the data.

Model Files

Trained model files such as:

.keras
.h5

are excluded from Git when they exceed practical repository limits.

This keeps the GitHub repository focused on the code, experiments, methodology, and documentation rather than large binary artifacts.

Goals

The main goals of VegHealth are to:

  1. Explore Deep Learning for agricultural computer vision.
  2. Build image-classification models for different vegetables.
  3. Compare model performance across vegetable health conditions.
  4. Develop reusable preprocessing and training workflows.
  5. Investigate how computer vision could support early plant-disease detection.
  6. Eventually integrate the individual models into a broader VegHealth system.

Development Roadmap

Phase 1 — Foundation

  • Set up project repository
  • Create vegetable-specific project structure
  • Begin tomato classification
  • Explore tomato dataset
  • Complete tomato model evaluation

Phase 2 — Expansion

  • Add potato classification
  • Add pepper classification
  • Add additional vegetables
  • Standardize preprocessing
  • Compare model performance

Phase 3 — Integration

  • Develop a unified inference pipeline
  • Accept an uploaded vegetable image
  • Identify the vegetable
  • Predict its health condition
  • Return prediction confidence
  • Build a simple user interface

Phase 4 — Improvement

  • Experiment with transfer learning
  • Improve generalization
  • Evaluate class imbalance
  • Perform error analysis
  • Test the system on unseen images

Project Philosophy

VegHealth is being developed as a learning and research project, with an emphasis on understanding the complete Deep Learning workflow rather than simply training a model and reporting its accuracy.

The project covers:

Data → Exploration → Preprocessing → Modeling → Training → Evaluation → Deployment

Disclaimer

VegHealth is an experimental computer vision project.

Model predictions should not be treated as a definitive agricultural diagnosis. Real-world deployment would require testing on representative field data and validation by agricultural experts.

Author

Ruvarashe S Nemaramba

BSc Honours Artificial Intelligence Africa University

GitHub: Ruvaaa

VegHealth is a work in progress.