Bio-Inspired Spotted Hyena Optimizer with Deep Convolutional Neural Network-Based Automated Food Image Classification

Area of Science:

• Computer Vision
• Deep Learning
• Artificial Intelligence

Background:

• Food image classification is crucial for applications like dietary monitoring and restaurant recommendations.
• Deep Learning (DL) and Convolutional Neural Networks (CNNs) have significantly advanced automated image classification.
• Existing methods require optimization for improved accuracy in complex food image datasets.

Purpose of the Study:

• To develop an advanced automated food image classification system.
• To enhance the accuracy and efficiency of food recognition from images.
• To introduce a novel hybrid approach combining DL with bio-inspired optimization.

Main Methods:

• A Deep Convolutional Neural Network (DCNN) utilizing the Xception model for feature extraction.
• Integration of the Spotted Hyena Optimizer (SHO) algorithm for hyperparameter tuning of the DCNN.
• Employing the Extreme Learning Machine (ELM) model for final food image classification.

Main Results:

• The proposed SHODCNN-FIC method demonstrated superior performance in food image classification compared to other DL models.
• Experimental results validated the effectiveness of the SHO algorithm in optimizing DCNN hyperparameters.
• Accurate classification of diverse food items was achieved, showcasing the model's robustness.

Conclusions:

• The SHODCNN-FIC approach offers a highly effective solution for automated food image classification.
• The synergy between DCNNs, SHO, and ELM significantly improves recognition accuracy.
• This method holds potential for various applications requiring precise food identification from visual data.