Food Image Classification

This repository contains a deep learning project that classifies images of 101 different foods. The project utilizes a pretrained MobileNetV2 model from TensorFlow/Keras and fine-tunes it with additional dense layers to achieve accurate food category predictions.

Overview

The objective of this project is to build an image classification system that identifies the food present in a given image. We employ a TensorFlow/Keras pretrained MobileNetV2 model, which is further fine-tuned with additional fully connected layers to classify images into 101 food categories. The project also includes steps for data preparation, train-test splitting, model training with early stopping, and evaluation using confusion matrices and classification reports.

Features

• Data Preparation:
  • Downloads the Food-41 dataset from Kaggle.
  • Unzips and organizes images into a structured DataFrame.
• Data Splitting & Augmentation:
  • Splits the dataset into training, validation, and testing sets.
  • Uses TensorFlow's ImageDataGenerator with MobileNetV2 preprocessing.
• Modeling & Transfer Learning:
  • Leverages MobileNetV2 as a feature extractor (with frozen weights).
  • Adds two Dense layers and a softmax output layer to classify 101 food categories.
• Training & Early Stopping:
  • Uses the Adam optimizer and categorical crossentropy loss.
  • Incorporates early stopping to prevent overfitting.
• Evaluation & Visualization:
  • Evaluates the model on a test set.
  • Generates confusion matrices and classification reports.
  • Visualizes the confusion matrix using Seaborn and Matplotlib.

Dataset link - https://www.kaggle.com/datasets/kmader/food41