Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Jun 26, 2026

Anesthesia-free Heartbeat Measurements in Freely Moving Zebrafish
03:57

Anesthesia-free Heartbeat Measurements in Freely Moving Zebrafish

Published on: April 18, 2025

CNN-Based Classification of Ziziphus Seeds with Focal Loss for Overcoming Size-Based Shortcut Learning.

Yea-Jin Park1,2, Dae-Hyun Jung1,2

  • 1Department of Smart Farm Science, Kyung Hee University, Yongin 17104, Republic of Korea.

Biosensors
|June 25, 2026
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Depth image conversion model based on CycleGAN for growing tomato truss identification.

Plant methodsยท2022
See all related articles

Counterfeit herbal medicines threaten food safety. This study found focal loss effectively improves AI model accuracy by focusing on intrinsic features, not size, for reliable authentication.

Area of Science:

  • Agricultural Science
  • Computer Science
  • Food Science

Background:

  • Counterfeit herbal medicines pose a significant threat to global food safety.
  • AI models often rely on superficial features (shortcut learning) instead of intrinsic characteristics for classification.
  • Distinguishing authentic *Ziziphus jujuba* Mill. var. *spinosa* from its counterfeit *Ziziphus mauritiana* Lam. is challenging due to morphological similarities.

Purpose of the Study:

  • To identify and mitigate size-based shortcut learning in AI models for herbal medicine authentication.
  • To evaluate the effectiveness of focal loss as a preprocessing-free method for improving model generalization.
  • To enhance the reliability of AI-driven herbal medicine authentication for field applications.

Main Methods:

Keywords:
deep learningexplainable artificial intelligence (XAI)food safetyherbal medicineimage classificationloss function optimizationshortcut learning

Related Experiment Videos

Last Updated: Jun 26, 2026

Anesthesia-free Heartbeat Measurements in Freely Moving Zebrafish
03:57

Anesthesia-free Heartbeat Measurements in Freely Moving Zebrafish

Published on: April 18, 2025

  • Developed and evaluated AI models using cross-entropy and focal loss on internal and external datasets.
  • Investigated the impact of data preprocessing techniques, including background removal and size normalization.
  • Utilized Grad-CAM++ for visualizing model attention and confirming feature focus.

    • Main Results:

    • Size-based shortcuts significantly degraded model performance on external datasets.
    • Size normalization partially improved generalization but focal loss achieved superior results without preprocessing.
    • The focal loss model achieved 90.88% accuracy on the external dataset, significantly reducing the generalization gap.

    Conclusions:

    • Focal loss effectively mitigates size-based shortcut learning in herbal medicine authentication.
    • A preprocessing-free focal loss approach enhances AI model reliability and generalization for field use.
    • This method offers a practical solution for authenticating herbal medicines against counterfeits.