Related Experiment Video
Updated: May 28, 2026
13:19
Deep Neural Networks for Image-Based Dietary Assessment
Published on: March 13, 2021
Food Origin Authenticity Using Deep Learning and Citizen Science: Bananas Case Study
Nikolaos Fragkos1, Yamine Bouzembrak1, Sara Wilhelmina Erasmus2
1Information Technology, Wageningen University, Wageningen University and Research, 6700 HB Wageningen, The Netherlands.
Foods (Basel, Switzerland)
|May 27, 2026
Summary
This study uses Artificial Intelligence (AI) and citizen science images to detect the country of origin for Cavendish bananas, offering a new method for food fraud detection.
Area of Science:
- Agricultural Science
- Computer Science
- Food Science
Background:
- Food fraud is a significant global issue, impacting consumer trust and economic integrity.
- Accurate origin verification of agricultural products like bananas is challenging but crucial for supply chain transparency.
- Existing methods for origin detection may not be sufficient for proactive fraud prevention at the cultivation stage.
Purpose of the Study:
- To develop and evaluate an Artificial Intelligence (AI)-based proof-of-concept for detecting the country of origin of Cavendish bananas (Musa spp.).
- To explore the utility of citizen science-generated imagery in conjunction with deep learning for agricultural product authentication.
- To establish a foundational AI pipeline for proactive food fraud detection at the cultivation level.
Main Methods:
- Utilized a dataset of 6000 citizen science images of Cavendish bananas sourced from iNaturalist.
- Employed Convolutional Neural Networks (CNNs), specifically leveraging transfer learning with pre-trained models like MobileNetV1.
- Implemented data augmentation techniques to address class imbalance and enhance model robustness.
Main Results:
- The CNN model, particularly MobileNetV1, showed promising performance in distinguishing banana origins across six countries.
- Achieved an average accuracy of 0.86 with Monte Carlo Cross Validation and 0.77 with 5-Fold Cross Validation.
- The final selected model demonstrated a validation accuracy of 0.79, indicating significant potential for origin detection.
Conclusions:
- This study presents a foundational proof-of-concept for AI-driven origin detection in agricultural products at the cultivation stage.
- The findings suggest a promising new avenue for proactive food fraud detection using accessible imagery and advanced computational models.
- The developed AI pipeline offers a scalable framework that can be further expanded and independently validated for real-world applications.
Related Concept Videos
01:29Sources of Food Contamination
Contamination of food by microbial agents and natural toxins poses significant risks to public health. These hazards can be introduced at various points across the food supply chain, ranging from environmental sources to processing and storage stages. Understanding these contamination pathways is critical for developing strategies to ensure food safety.Seafood is particularly vulnerable to contamination through both environmental exposure and microbial colonization. Toxins from harmful algal...
02:35Key Elements for Plant Nutrition
Like all living organisms, plants require organic and inorganic nutrients to survive, reproduce, grow and maintain homeostasis. To identify nutrients that are essential for plant functioning, researchers have leveraged a technique called hydroponics. In hydroponic culture systems, plants are grown—without soil—in water-based solutions containing nutrients. At least 17 nutrients have been identified as essential elements required by plants. Plants acquire these elements from the atmosphere, the...