This project is a deep learning-based food image classification system designed to recognize food items and estimate their nutritional facts. The model leverages transfer learning with the DenseNet201 architecture, pre-trained on ImageNet, to classify images into 101 food categories.

• Classification of 101 different food categories.
• Robust data augmentation to improve model generalization.
• Transfer learning using DenseNet201 for efficient training.
• Interactive visualizations of training and validation performance.

The dataset used in this project is the Food-101 dataset. It is a large dataset of food images organized by food type and includes:

• Size: 101 food categories with 750 training and 250 test images per category, making a total of 101,000 images.
• Quality: Labels for the test images have been manually cleaned, while the training set contains some noise.
• Context: This dataset is ideal for testing computer vision techniques and is also available on the ETH Zurich website.
• Source: Detailed in the paper "Food-101 – Mining Discriminative Components with Random Forests" by Lukas Bossard, Matthieu Guillaumin, and Luc Van Gool.

1. Clone the repository:

   git clone https://github.com/KenanSh/Food-Recognition-and-Nutrition-Facts-Estimation-Fixed-Values

2. Install required dependencies:

   pip install -r environment.yaml

This project has demonstrated the potential of deep learning for food classification but also highlighted key limitations:

• Accuracy Challenges: The classification accuracy is limited (48%) due to the large number of classes and insufficient samples per class.
• Inefficient Nutritional Estimation: The current method relies on static information for each food class, which is not scalable or practical.
• Planned Improvements:
  • Develop a solution to estimate the weight of food items in images.
  • Build or source a dataset specifically designed for nutritional facts estimation.