CLAPE: A Validated Multimodal Dataset for Physiological Student Engagement Prediction Using Low-Cost RGB Webcam-Based Pupil Variation and Contextual Learner Characteristics
The CLAPE (Contextual Learner Attributes and Pupil Engagement) dataset is a validated multimodal educational dataset designed to support research on physiological student engagement using low-cost, non-invasive RGB webcam technology. The dataset integrates physiological pupil variation measurements with demographic, academic, anthropometric, behavioural, lifestyle, and cultural learner characteristics collected under standardized educational settings. The dataset comprises 272 unique undergraduate participants enrolled in Computer Science and related programmes at Amity University Punjab, India. Data collection was conducted through eight standardized laboratory-based experimental sessions, with 34 participants attending each session, resulting in a total of 272 unique participant records. Each participant contributed data during a single experimental session. All sessions followed an identical experimental protocol to ensure consistency and reproducibility. Four sessions were conducted in English and four in Hindi, with all participants receiving the same 30-minute Database Management Systems (DBMS) lecture delivered by the same instructor under controlled indoor lighting conditions. Before each lecture, a 60-second baseline pupil recording was obtained, followed by continuous real-time pupil monitoring using a minimum 5 MP RGB webcam. Physiological data acquisition was performed using MediaPipe Face Mesh and OpenCV-based pupil localization, integrated within a customized Student Experiment Portal, while a secure Teacher Dashboard stored and managed raw and processed experimental data. The dataset includes comprehensive learner metadata, encompassing demographic information, academic performance, height, weight, body mass index (BMI), parental education, newspaper reading frequency, religious reading habits, worship frequency, cultural participation, sleep duration, breakfast status, exercise habits, vision-related information, computer learning interest, and other contextual learner characteristics. Additionally, the dataset provides validated engineered indices, including the Academic Score History (ASH), Awareness Exposure Index (AEI), Cultural Expression Index (CEI), Ritual Participation Index (RPI), and Contextual Influence Score (CIS). Physiological engagement is operationalized using participant-specific baseline-normalized pupil variation. A participant is classified as physiologically engaged when pupil variation remains at least 1.25% above the individualized baseline threshold for 12 minutes or more during the 30-minute lecture. The engagement criterion was established through statistical validation and repeated experimental analyses to ensure reproducibility and interpretability. The repository includes the processed dataset, metadata, variable dictionary, participant questionnaire, experimental protocol, ethics approval documentation, benchmark machine learning results, Student Experiment Portal, Teacher Dashboard, and supporting documentation required for reproducible research. The CLAPE dataset is intended to support research in educational data mining, learning analytics, physiological computing, multimodal learning analytics, computer vision, explainable artificial intelligence, human–computer interaction, and machine learning. It provides an openly accessible benchmark resource for developing, validating, benchmarking, and comparing computational methods for physiological student engagement prediction while promoting transparency, reproducibility, and reuse within the educational AI research community. This version correctly reflects the study design of 272 unique participants distributed across eight laboratory sessions, rather than repeated observations from the same 34 participants. All participant information has been fully anonymized to protect privacy and confidentiality. The dataset contains no personally identifiable information. By downloading or using this dataset, users agree to comply with the accompanying Data Usage Agreement (DUA) and to use the dataset solely for lawful research and educational purposes.