Application of Deep Learning in Food: A Review

Area of Science:

• Food Science and Engineering
• Artificial Intelligence
• Data Science

Background:

• Deep learning (DL) is an advanced big data analysis technology with proven success in various fields.
• Its application in food science and engineering is a recent development, with limited comprehensive reviews available.
• This review is the first to focus on DL applications within the food domain.

Purpose of the Study:

• To provide an overview of deep learning architectures and training methodologies.
• To survey existing research applying DL in food science and engineering.
• To analyze DL's potential for food sensory and consumer research.

Main Methods:

• A comprehensive literature survey of articles utilizing DL for food-related problems.
• Detailed investigation of specific problems, datasets, preprocessing, networks, frameworks, and performance metrics.
• Comparison of DL methods with conventional machine learning algorithms and manual feature extraction techniques.

Main Results:

• Deep learning models demonstrated superior performance over traditional methods in food recognition, quality detection, and contamination assessment.
• DL achieved encouraging results in classification and regression tasks across various food products (fruits, vegetables, meat, aquatic products).
• The study identified DL as a highly promising tool for food quality and safety inspection.

Conclusions:

• Deep learning is a powerful and effective tool for addressing challenges in food science and engineering.
• Its application in food quality and safety inspection significantly outperforms existing methods.
• Future research is expected to increasingly leverage DL for advanced data mining in food sensory and consumer studies.