https://github.com/abhayy-kumar/food-image-classification

A deep learning technique that leverages transfer learning to classify images into different food categories.
https://github.com/abhayy-kumar/food-image-classification

deep-learning image-classification mobilenetv2 tensorflow trasfer-learning

Last synced: 3 months ago
JSON representation

A deep learning technique that leverages transfer learning to classify images into different food categories.

• Host: GitHub
• URL: https://github.com/abhayy-kumar/food-image-classification
• Owner: Abhayy-Kumar
• Created: 2025-03-21T05:18:17.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2025-03-21T05:22:12.000Z (over 1 year ago)
• Last Synced: 2025-06-02T00:29:35.922Z (about 1 year ago)
• Topics: deep-learning, image-classification, mobilenetv2, tensorflow, trasfer-learning
• Language: Jupyter Notebook
• Homepage:
• Size: 512 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# 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