Datasets
Datasets
In our lab, we collect and publish datasets to support reproducible research and support further investigation of human behavior, communication and interaction.
Below, you can find an overview of our available datasets and links to the corresponding Zenodo repositories, including information on how to receive access.
- CHILL - Challenging Heartrate and Illumination rPPG
The CHILL dataset supports research on camera-based heart-rate estimation using remote photoplethysmography (rPPG) under challenging conditions. It contains preprocessed facial video frames and corresponding heart-rate labels across four scenarios combining low vs. high heart rate with bright vs. low-light conditions. Find it on Zenodo here.
Bhargav Acharya, William Saakyan, Prof. Dr. Barbara Hammer, Prof. Dr. Hanna Drimalla
- CUES - Confusion and Understandung: Expressions Under Stress
The CUES dataset provides facial-behavior features from participants who received a board-game explanation under both stressful and neutral conditions. Facial features were extracted using OpenFace, while participants continuously self-annotated their experienced understanding and confusion. The dataset can support research on behavioral expressions of cognitive states and how these expressions change under stress. Find it on Zenodo here.
Jonas Paletschek, Prof. Dr. Hanna Drimalla, Dr. David Johnson
- Dona: De-identified Messaging Meta-Data
De-identified Messaging Data (WhatsApp and Facebook) and Evaluation Responses
This dataset contains de-identified messaging metadata from WhatsApp and Facebook data donations collected with our tool Dona. In addition to messaging behavior, it includes sociodemographic information as well as participants’ evaluations of personalized visual feedback concerning their own communication patterns, thus enabling research on digital communication as well as factors of the data donation process itself. Find it on Zenodo here.
Dr. Olya Hakobyan, Paul-Julius Hillmann, Prof. Dr. Hanna Drimalla
Comparison of objective WhatsApp data and subjective self-reports before and after data-driven personalized feedback
This dataset combines de-identified WhatsApp messaging metadata collected through our platform Dona with self-reported communication behavior before and after participants received personalized, data-driven visual feedback on their own communication patterns. It enables comparisons between objectively measured messaging patterns and users’ perceptions of their own behavior as well as investigations of whether personalized feedback influences self-assessment. Find it on Zenodo here.
Dr. Olya Hakobyan, Paul-Julius Hillmann, Prof. Dr. Hanna Drimalla
WhatsApp and Instagram Chat Message Metadata (WICM)
The WICM dataset contains privacy-preserving metadata from donated WhatsApp and Instagram conversations collected with our tool Dona, including timestamps, message-length information and anonymized identifiers without any of the original message content. The data enable large-scale analyses of temporal communication patterns, such as response speeds, reciprocity between conversation partners or stability of these interaction patterns over time. Find it on Zenodo here.
Florian Martin, Dr. Olya Hakobyan, Prof. Dr. Hanna Drimalla