Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Summary of the Unusual Activity Recognition Challenge for Developmental Disability Support

Christina Garcia, Nhat Tan Le, Taihei Fujioka, Umang Dobhal, Milyun Ni'ma Shoumi, Thanh Nha Nguyen, Sozo Inoue

TL;DR

ISAS 2025 tackles unusual activity recognition for developmental disability support using pose-based data. It introduces an eight-class dataset (four normal, four unusual) with two evaluation phases: Task 1 on an unseen subject and Task 2 based on a five-subject LOSO protocol, using macro F1 as the primary metric. Results show that hybrid ML/DL ensembles and temporal models can achieve strong performance, but generalization to unseen individuals remains challenging, especially for abrupt, rare actions in low-dimensional skeleton data. The study emphasizes privacy-preserving, deployment-considerate HAR for healthcare and offers practical insights on windowing, imbalance handling, and robust temporal modeling for real-world care settings.

Abstract

This paper presents an overview of the Recognize the Unseen: Unusual Behavior Recognition from Pose Data Challenge, hosted at ISAS 2025. The challenge aims to address the critical need for automated recognition of unusual behaviors in facilities for individuals with developmental disabilities using non-invasive pose estimation data. Participating teams were tasked with distinguishing between normal and unusual activities based on skeleton keypoints extracted from video recordings of simulated scenarios. The dataset reflects real-world imbalance and temporal irregularities in behavior, and the evaluation adopted a Leave-One-Subject-Out (LOSO) strategy to ensure subject-agnostic generalization. The challenge attracted broad participation from 40 teams applying diverse approaches ranging from classical machine learning to deep learning architectures. Submissions were assessed primarily using macro-averaged F1 scores to account for class imbalance. The results highlight the difficulty of modeling rare, abrupt actions in noisy, low-dimensional data, and emphasize the importance of capturing both temporal and contextual nuances in behavior modeling. Insights from this challenge may contribute to future developments in socially responsible AI applications for healthcare and behavior monitoring.

Summary of the Unusual Activity Recognition Challenge for Developmental Disability Support

TL;DR

ISAS 2025 tackles unusual activity recognition for developmental disability support using pose-based data. It introduces an eight-class dataset (four normal, four unusual) with two evaluation phases: Task 1 on an unseen subject and Task 2 based on a five-subject LOSO protocol, using macro F1 as the primary metric. Results show that hybrid ML/DL ensembles and temporal models can achieve strong performance, but generalization to unseen individuals remains challenging, especially for abrupt, rare actions in low-dimensional skeleton data. The study emphasizes privacy-preserving, deployment-considerate HAR for healthcare and offers practical insights on windowing, imbalance handling, and robust temporal modeling for real-world care settings.

Abstract

This paper presents an overview of the Recognize the Unseen: Unusual Behavior Recognition from Pose Data Challenge, hosted at ISAS 2025. The challenge aims to address the critical need for automated recognition of unusual behaviors in facilities for individuals with developmental disabilities using non-invasive pose estimation data. Participating teams were tasked with distinguishing between normal and unusual activities based on skeleton keypoints extracted from video recordings of simulated scenarios. The dataset reflects real-world imbalance and temporal irregularities in behavior, and the evaluation adopted a Leave-One-Subject-Out (LOSO) strategy to ensure subject-agnostic generalization. The challenge attracted broad participation from 40 teams applying diverse approaches ranging from classical machine learning to deep learning architectures. Submissions were assessed primarily using macro-averaged F1 scores to account for class imbalance. The results highlight the difficulty of modeling rare, abrupt actions in noisy, low-dimensional data, and emphasize the importance of capturing both temporal and contextual nuances in behavior modeling. Insights from this challenge may contribute to future developments in socially responsible AI applications for healthcare and behavior monitoring.
Paper Structure (17 sections, 7 figures, 3 tables)

This paper contains 17 sections, 7 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Irregular and Unpredictable Activities in HAR
  4. Data Imbalance between Normal versus Unusual Activities
  5. Temporal and Subject Variability
  6. Language Models in Human Activity Recognition
  7. Details of the Challenge
  8. Challenge Dataset
  9. Limitations in the Data
  10. Goals of the Challenge
  11. Evaluation Protocol
  12. Results of the Challenge
  13. Participating Teams
  14. Task 1: Predicting Unseen 1 Person Test Data
  15. Performance with 5 Participant Leave One Subject Out
  16. ...and 2 more sections

Figures (7)

  • Figure 1: Challenge Overview Dataset and with Two Tasks
  • Figure 2: Examples of unusual behaviors represented in the dataset using pose estimation
  • Figure 3: Team size distribution across participating groups (left) and challenge completion rate (right)
  • Figure 4: Country-wise distribution of participants. Vietnam had the highest number of participants, indicating strong regional engagement. Meanwhile, countries like Poland and the United Kingdom had the least number of participants, with only one representative each.
  • Figure 5: Task 1 Model Performance Comparison, One Unseen Person Test
  • ...and 2 more figures