🌾 Crops Sorter Model

A machine learning pipeline for detecting crop health patterns using Convolutional Neural Networks (CNNs). The system is trained on 10,000+ labeled agricultural images to predict crop conditions with high accuracy, and includes analytics and visualizations of model performance.

---

🧠 Core Features

• 🤖 Trained 3 custom CNN architectures using TensorFlow and Keras
• 📊 Achieved 90.07% accuracy on validation data
• 📈 Designed analytics scripts to evaluate 4 key model metrics:
  • Accuracy
  • Precision
  • Recall
  • Loss
• 🌐 Visualized predictions using Matplotlib with bar graphs and cluster charts

---

🛠️ Tech Stack

Purpose	Technology
Model Training	TensorFlow, Keras
Data Preprocessing	Scikit-learn
Data Visualization	Matplotlib
Language	Python

---

🧪 Training Overview

• Dataset: >10,000 images across various crop-health categories
• Preprocessing: Normalization, augmentation, and one-hot encoding
• CNN Variants:
  • ResNet-inspired shallow net
  • Custom-built 6-layer CNN
  • Lightweight MobileNetv2 baseline
• Evaluation: Accuracy calculated via validation split (90.07%)

---

📊 Visualizations

Using matplotlib, the following charts were generated:

• Bar graphs comparing model performance
• Cluster plots to group prediction categories
• Accuracy/Loss trends over training epochs

---

📁 Files

• train_model.py — Model architecture and training loop
• evaluate.py — Script to calculate and compare model metrics
• visualize.py — Generates performance graphs
• README.md — Project documentation

---

🧠 Future Improvements

• Integrate early stopping and learning rate schedulers
• Expand dataset with underrepresented crops
• Deploy trained model to a web/mobile interface for farmer use

---

📅 Timeline

January – February 2025
Created as a solo AI-agriculture capstone to explore the intersection of deep learning and food security.

---

📬 Contact

Kevin Chifamba
📧 kevinnanashe@gmail.com
🔗 LinkedIn • GitHub

---