Food Recognizer

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference Check out for more resources: https://huggingface.co/spaces/tdnathmlenthusiast/food_classifier

Food Classifier Project

Overview

The Food Classifier project is an innovative deep learning-based application that aims to accurately identify and categorize various types of food items from images. This project leverages the power of state-of-the-art machine learning techniques to provide users with an intuitive and efficient way to classify foods, whether for dietary monitoring, restaurant menu analysis, or personal cooking exploration. By combining computer vision and neural networks, the Food Classifier brings automation and convenience to the world of food recognition.

Introduction

The Food Classifier project addresses the need for a reliable and efficient tool to identify and categorize various food items from images. With the increasing interest in healthy eating, dietary restrictions, and culinary exploration, this project serves as a valuable asset for both individuals and businesses. The application of deep learning techniques to this domain not only simplifies food identification but also opens doors to personalized nutritional analysis and recipe recommendations.

Motivation

The motivation behind the Food Classifier project is to bridge the gap between cutting-edge machine learning technologies and everyday scenarios. We live in a visual world, and food is an integral part of our lives. Imagine being able to take a photo of a meal and instantly knowing its nutritional value or being presented with similar recipes to try. This project strives to make such experiences possible by harnessing the power of convolutional neural networks (CNNs) and transfer learning.

Features

The Food Classifier project boasts a range of features that make it stand out:

Multi-class Classification: The model can classify a diverse range of food items into multiple categories.
User-friendly Interface: The user interface is intuitive, allowing users to easily upload images for classification.
Nutritional Information: The application provides basic nutritional information about the recognized food items.
Recipe Recommendations: Users can receive recipe suggestions based on the classified food item, opening doors to culinary exploration.
Scalability: The project is designed to handle a growing food database, accommodating new and rare food items.

Data Collection

Clone the repo

git clone https://github.com/darkangrycoder/food_classifier_model.git

Intialize and activate virtual environment

virtualenv --no-site-packages  venv
source venv/bin/activate

Install dependencies
```
pip install -r requirements.txt
```
Download Chrome WebDrive from https://chromedriver.chromium.org/downloads
Run the scrapers
```
python -u food_item_scraping-1.py 
```

Data Preprocessing

Raw image data can be noisy and inconsistent. To ensure reliable model training, the dataset underwent preprocessing steps such as resizing, normalization, and augmentation. Data augmentation techniques, including rotation, flipping, and brightness adjustment, were applied to enhance the model’s robustness and ability to handle variations in image quality.

Numbers of Sample

Before Cleaning: 6000+ After Cleaning: 5900

Model Architecture

The core of the Food Classifier project lies in its deep learning model architecture. A pre-trained CNN architecture, such as VGG16, ResNet, or Inception, forms the backbone of the model. Transfer learning enables the project to leverage the knowledge learned from large datasets in other domains, enhancing the model’s accuracy even with limited food-specific training data.

Training

Training the model involves fine-tuning the pre-trained architecture on the curated food dataset. The model learns to recognize distinctive features of different food items, enabling it to differentiate between visually similar dishes. The training process also involves fine-tuning hyperparameters, such as learning rate and batch size, to achieve optimal performance.

Evaluation

The model’s performance is rigorously evaluated using various metrics such as accuracy, precision, recall, and F1 score. A separate validation dataset, distinct from the training data, is used to assess the model’s generalization capability. The evaluation results guide further iterations and improvements in the model.

Usage

Using the Food Classifier is as simple as uploading an image through the user interface. Once the image is uploaded, the model processes it and provides predictions about the food item’s category. Users can explore nutritional information and receive recipe recommendations based on the recognized food item.

Future Enhancements

The Food Classifier project is an ongoing endeavor with potential future enhancements, including:

Fine-grained Classification: Enhancing the model to distinguish between different variations of the same dish.
Mobile Application: Developing a mobile app for real-time food classification on the go.
User Customization: Allowing users to add new food categories and contribute to the project’s dataset.
Localization: Expanding the project to recognize food items from different cuisines and cultures.

Contributing

Contributions to the Food Classifier project are welcome! Whether through code improvements, dataset expansion, or feature suggestions, the project benefits from a collaborative community effort. Please refer to the Contributing Guidelines for more information on how to get involved.

License

This project is licensed under the MIT License, allowing for open collaboration and derivative works.

The Food Classifier project showcases the potential of combining deep learning, computer vision, and culinary exploration. By accurately classifying food items and providing insightful information, this project empowers users to make informed dietary choices and embark on delightful gastronomic adventures.