FoodSenseAI

• The project is built upon the principles of deep learning and incorporates advanced techniques to classify a dataset containing 16,643 food images grouped into 11 categories

• Dataset from https://www.kaggle.com/datasets/vermaavi/food11/data

• Built while learning from https://www.coursera.org/projects/transfer-learning-food-classification

• Skills -> Computer Vision | Image Classification | Python | Tensorflow | Keras | CNN | Transfer Learning | PreTrained Models

• Implementation -> https://github.com/TechWithRamaa/FoodSenseAI/blob/main/food_sense_ai.ipynb

Data Handling and Preparation:

• Utilizes Pandas and NumPy for efficient data manipulation
• Real-time data augmentation is achieved using ImageDataGenerator to improve model generalization

Model Architecture

• Employs Convolutional Neural Networks (CNNs) for feature extraction from food images
• Integrates pre-trained model InceptionResNetV2 through transfer learning to leverage learned features, enhancing classification accuracy

Optimization and Training:

• Uses Stochastic Gradient Descent (SGD) and other advanced optimizers for effective learning
• Implements callbacks to manage learning rates, prevent overfitting, and save the best model versions

Visualization and Analysis:

• Employs Matplotlib and Seaborn for visualizing data distributions, model performance, and training metrics