Fruit Classification Using Transfer Learning

Project Overview

This project focuses on the development of a fruit classification system using transfer learning with the pre-trained VGG16 model. The model was fine-tuned on a custom fruit image dataset to classify different fruit categories while minimizing data and computation requirements.

The system achieves high accuracy by leveraging learned ImageNet features and retraining only the top layers. The project includes performance evaluation, accuracy visualization, and prediction testing on sample images.

Image Gallery

Sample predictions and classification results:

Key Features

• Transfer learning using VGG16
• Fine-tuned on custom fruit dataset
• High accuracy with minimal training data
• Visualization of predictions on test images
• Efficient classification of multiple fruit categories

Technologies Used

Python TensorFlow / Keras NumPy Transfer Learning VGG16 Matplotlib

Model Development Workflow

• Dataset Preparation: Load, resize, and normalize fruit images.
• Transfer Learning Setup: Freeze VGG16 base layers.
• Fine-Tuning: Train new fully connected layers for fruit classes.
• Evaluation: Calculate accuracy and loss curves.
• Prediction Visualization: Display model predictions on test images.

Sample Code Snippet

from tensorflow.keras.applications import VGG16
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Flatten
from tensorflow.keras.preprocessing.image import ImageDataGenerator

base_model = VGG16(weights='imagenet', include_top=False, input_shape=(224,224,3))
base_model.trainable = False   # Freeze convolutional layers

model = Sequential([
    base_model,
    Flatten(),
    Dense(256, activation='relu'),
    Dense(5, activation='softmax')  # Example: 5 fruit classes
])

model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
      

Project Outcome

The transfer learning model achieved strong accuracy while using a small dataset. Its ability to generalize well across fruit categories makes it suitable for educational purposes and lightweight real-world applications.