Related Experiment Video
Updated: Jan 15, 2026
13:19
Deep Neural Networks for Image-Based Dietary Assessment
Published on: March 13, 2021
9.9K
Deep Learning-Based Image Recognition for Food Science and Technology: End-to-End Workflows and Domain-Specific
Bin Liao1, Yiming Wang2, Xiang Li3,4
1College of Food Science, Southwest University, Chongqing, China.
Comprehensive Reviews in Food Science and Food Safety
|January 14, 2026
Summary
Deep learning for food science offers intelligent systems for quality and safety but faces challenges. This review proposes a workflow and strategies to improve data, models, and integration for reliable computer vision in food technology.
Area of Science:
- Food Science
- Computer Vision
- Artificial Intelligence
Background:
- Deep learning image recognition is transforming food science applications like grading and safety.
- Adoption is limited by fragmented practices, lack of guidance, and inconsistent methods.
Purpose of the Study:
- To synthesize a modular, end-to-end workflow for deep learning in food science.
- To address challenges like data scarcity, annotation issues, and submodel integration.
Main Methods:
- Review of current practices and challenges in food image recognition.
- Development of domain-specific strategies for workflow components.
- Examination of three case studies to illustrate applications and hurdles.
Main Results:
- A comprehensive workflow from task formulation to application is presented.
- Domain-specific strategies are proposed for data refinement, annotation, preprocessing, augmentation, and integration.
- Common challenges including limited data and integration obstacles are highlighted.
Conclusions:
- Standardized toolkits and platforms are needed for reliable, generalizable computer vision in food science.
- Addressing current challenges will enhance reproducibility and scalability of AI solutions.
- Future efforts should focus on unified toolchains and integration platforms for real-world impact.
