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: Jul 4, 2026

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
04:23

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images

Published on: April 21, 2023

Cheatomaly: weakly supervised video anomaly ranking for exam cheating detection using vision transformers.

El Mehdi Alaoui Mrani1, Anas Bouayad1, Khalid Fardousse2

  • 1LIASSE Laboratory, ENSA, SMBA University, Fes, Morocco.

Frontiers in Big Data
|May 28, 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

Compact waste image classification with multi-student CNNs and edge-oriented model selection.

Frontiers in artificial intelligenceยท2026
See all related articles

Detecting subtle classroom cheating is hard. We introduce Cheatomaly, a video dataset and weakly supervised method using Multiple Instance Learning (MIL) to identify suspicious behaviors, creating a new benchmark for exam integrity.

Area of Science:

  • Computer Science
  • Artificial Intelligence
  • Machine Learning

Background:

  • Detecting cheating in classroom exams is difficult due to subtle, sparse, and context-dependent behaviors.
  • Existing benchmarks for anomaly detection do not adequately address the nuances of academic dishonesty in educational settings.

Purpose of the Study:

  • To introduce Cheatomaly, a novel video dataset for detecting cheating in classroom examinations.
  • To develop and evaluate a weakly supervised anomaly detection framework for identifying suspicious behaviors during exams.

Main Methods:

  • Formulated cheating detection as a weakly supervised video anomaly ranking task using Multiple Instance Learning (MIL).
  • Utilized Vision Transformer features and a Mean, Standard Deviation, and Temporal Difference (MSD) formulation for segment-level representations.
Keywords:
anomaly detectioncheating detectionmultiple instance learningtemporal modelingvideo analysisvision transformersweakly supervised learning

Related Experiment Videos

Last Updated: Jul 4, 2026

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
04:23

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images

Published on: April 21, 2023

  • Employed a margin-based ranking objective with video-level labels for training.

    • Main Results:

    • Achieved strong video-level discrimination and meaningful frame-level localization on the Cheatomaly dataset.
    • Demonstrated that temporal aggregation strategies impact the trade-off between ranking and localization performance.
    • Identified context-dependent temporal behavior modeling as the primary challenge, rather than feature aggregation alone.

    Conclusions:

    • Cheatomaly serves as a realistic benchmark for studying subtle cheating anomalies in educational environments.
    • The proposed MIL approach shows promise for weakly supervised detection of exam misconduct.
    • Future research should focus on effectively modeling context-dependent temporal dynamics for improved cheating detection.