Fruit Image Classification

Overview

This project implements an advanced image classification solution using a pretrained MobileNetV2 model to accurately classify images of fruits. The solution encompasses comprehensive data preprocessing, model training, and prediction capabilities.

Project Structure

fruit-image-classification/
│
├── data/
│   ├── archive (6).zip
│   └── MY_data
|     ├── train/           # Training images
│     ├── test/            # Testing images
│     └── prediction/      # Images for prediction
│
├── src/
│   ├── preprocessing.py     # Dataset preprocessing and visualization
│   ├── model_training.py    # Model training script
│   ├── classify.py          # Prediction script
│   └── config.py            # Centralized configuration
│
├── results/              # Output storage
|   ├── plots/
|   |    └── sample_plot.png
│   ├── model_architecture.json
|   └── classifier_history.json
│
├── trained models/
│   └── fruit_classifier_model.h5
│
├── requirements.txt
└── README.md

Key Features

• Pretrained Model: Leverages MobileNetV2 for robust fruit image classification
• Advanced Preprocessing: Comprehensive image normalization
• Training Optimization:
  • Early stopping callback
  • Performance monitoring
• Visualization Tools: Dataset sample plotting
• Flexible Prediction: Real-time image classification script

Installation

Prerequisites

• Python 3.8+
• pip package manager

Setup Steps

1. Clone the Repository

   git clone https://github.com/Nirikshan95/FruitClassifier.git
   cd FruitClassifier

2. Install Dependencies

   pip install -r requirements.txt

3. Prepare Dataset

   • Download "Fruit Classification (10 Class)" dataset from Kaggle
   • Place archive (6).zip in the data/ directory
   • The preprocessing script will handle extraction

Usage

1. Train the Model

python src/model_training.py

2. Predict New Images

python src/classify.py

Note: Update demo_img path in classify.py with your image

3. Visualize Dataset

python src/preprocessing.py

Configuration

The config.py file centralizes project configurations:

# Paths
DATA_DIR = "./data"
ZIP_FILE_PATH = "./data/archive (6).zip"
TRAINING_DATA = "./data/MY_data/train/"
TESTING_DATA = "./data/MY_data/test/"
PREDICTION_DATA = "./data/MY_data/prediction/"
MODEL_SAVE_PATH = "./trained models/fruit_classifier_model.h5"
MODEL_HISTORY_PATH = "./results/classifier_history.json"
MODEL_ARCHITECTURE_PATH="./results/model_architecture.json"
SAMPLE_PLOT_PATH="./results/plots/sample_plot.png"

# Hyperparameters
BATCH_SIZE = 32
EPOCHS = 10
OPTIMIZER = "adam"

# Model Parameters
INPUT_SHAPE = (224,224,3)
ACTIVATION_FUNCTION = "relu"

Results

• Trained Model: fruit_classifier_model.h5
• Model Architecture: results/model_architecture.json
• Performance Metrics:
  • Training accuracy
  • Validation accuracy

Future Roadmap

• Expand fruit class diversity
• Fine-tune model architecture
• Implement advanced data augmentation
• Add model interpretability features

References

1. Kaggle Fruit Classification Dataset
2. MobileNetV2 Research Paper

Contributions

Contributions, issues, and feature requests are welcome!