# OPEN: A Benchmark Dataset and Baseline for Older Adult Patient Engagement Recognition in Virtual Rehabilitation Learning Environments

arXiv: https://arxiv.org/abs/2507.17959 · HTML: https://mlsanity.com/paper/2507.17959/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 17 of its 91 references that are in the map.
- First posted on arXiv 23 Jul 2025.

## Builds on (most cited first, 17 of 17)

- [Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition](https://mlsanity.com/paper/1801.07455/) (2018)
- [MediaPipe: A Framework for Building Perception Pipelines](https://mlsanity.com/paper/1906.08172/) (2019)
- [ROCKET: Exceptionally fast and accurate time series classification using random convolutional kernels](https://mlsanity.com/paper/1910.13051/) (2019)
- [MARLIN: Masked Autoencoder for facial video Representation LearnINg](https://mlsanity.com/paper/2211.06627/) (2022)
- [DAiSEE: Towards User Engagement Recognition in the Wild](https://mlsanity.com/paper/1609.01885/) (2016)
- [Prediction and Localization of Student Engagement in the Wild](https://mlsanity.com/paper/1804.00858/) (2018)
- [Attention Mesh: High-fidelity Face Mesh Prediction in Real-time](https://mlsanity.com/paper/2006.10962/) (2020)
- [Improving state-of-the-art in Detecting Student Engagement with Resnet and TCN Hybrid Network](https://mlsanity.com/paper/2104.10122/) (2021)
- [Do I Have Your Attention: A Large Scale Engagement Prediction Dataset and Baselines](https://mlsanity.com/paper/2302.00431/) (2023)
- [Graph-based Facial Affect Analysis: A Review](https://mlsanity.com/paper/2103.15599/) (2021)
- [Affect-driven Ordinal Engagement Measurement from Video](https://mlsanity.com/paper/2106.10882/) (2021)
- [Class-attention Video Transformer for Engagement Intensity Prediction](https://mlsanity.com/paper/2208.07216/) (2022)
- [Engagement Measurement Based on Facial Landmarks and Spatial-Temporal Graph Convolutional Networks](https://mlsanity.com/paper/2403.17175/) (2024)
- [TCCT-Net: Two-Stream Network Architecture for Fast and Efficient Engagement Estimation via Behavioral Feature Signals](https://mlsanity.com/paper/2404.09474/) (2024)
- [Detecting Disengagement in Virtual Learning as an Anomaly using Temporal Convolutional Network Autoencoder](https://mlsanity.com/paper/2211.06870/) (2022)
- [Bag of States: A Non-sequential Approach to Video-based Engagement Measurement](https://mlsanity.com/paper/2301.06730/) (2023)
- [Unobtrusive and Multimodal Approach for Behavioral Engagement Detection of Students](https://mlsanity.com/paper/1901.05835/) (2019)

## Abstract (arXiv)

Engagement in virtual learning is essential for participant satisfaction, performance, and adherence, particularly in online education and virtual rehabilitation, where interactive communication plays a key role. Yet, accurately measuring engagement in virtual group settings remains a challenge. There is increasing interest in using artificial intelligence (AI) for large-scale, real-world, automated engagement recognition. While engagement has been widely studied in younger academic populations, research and datasets focused on older adults in virtual and telehealth learning settings remain limited. Existing methods often neglect contextual relevance and the longitudinal nature of engagement across sessions. This paper introduces OPEN (Older adult Patient ENgagement), a novel dataset supporting AI-driven engagement recognition. It was collected from eleven older adults participating in weekly virtual group learning sessions over six weeks as part of cardiac rehabilitation, producing over 35 hours of data, making it the largest dataset of its kind. To protect privacy, raw video is withheld; instead, the released data include facial, hand, and body joint landmarks, along with affective and behavioral features extracted from video. Annotations include binary engagement states, affective and behavioral labels, and context-type indicators, such as whether the instructor addressed the group or an individual. The dataset offers versions with 5-, 10-, 30-second, and variable-length samples. To demonstrate utility, multiple machine learning and deep learning models were trained, achieving engagement recognition accuracy of up to 81 percent. OPEN provides a scalable foundation for personalized engagement modeling in aging populations and contributes to broader engagement recognition research.