SynPAIN

<h1>Overview</h1>
<p>SynPAIN is a publicly available synthetic dataset to support pairwise pain detection models for older adults with dementia. It contains 10,710 facial expression images (5,355 neutral/expressive pairs) across five ethnicities/races, representing two age groups (young: 20-35, old: 75+), both genders, and two expression types (pain and non-pain). It also includes five-second, 24 fps videos transitioning from a neutral to an expressive face for 40 identities, representing one combination from each ethnicity/race, gender, expression type, and age group.</p>
<p>Please check the metadata (under alternative URL) for a link to the <a href="https://huggingface.co/datasets/TaatiTeam/SynPAIN">HuggingFace repository</a> of this dataset.

<h2>Dataset Contents</h2>
<p>SynPAIN consists of the following:</p>
<ol>
<li>SynPain_Part1.zip: A zip file containing:
<ol>
<li>SynPain.txt: A txt file explaining the data structure and filename format.<\li>
<li>SynPain_folds.txt: A txt file providing the folds that were used for the "within-dataset experiments" (Section IV of the SynPAIN paper)</li>
<li>Images_Part1: A directory that contains 2,677 of the 5,355 images<\li>
<li>Videos: A directory that contains 40 videos, and for each video, the images they're based on<\li>
</ol></li>
<li>SynPain_Part1.zip: A zip file containing:
<ol>
<li>Images_Part2: A directory that contains 2,678 of the 5,355 images<\li>
</ol></li>
</ol>

<h2>Citation</h2>
<p>If you use SynPAIN in your research, please cite the SynPAIN paper:</p>
<p>Taati, B., Muzammil, M., Zarghami, Y., Moturu, A., Kazerouni, A., Mihailidis, A., Reimer, H., & Hadjistavropoulos, T. (2025). <a href="https://arxiv.org/abs/2507.19673">SynPAIN: A Synthetic Dataset of Pain and Non-Pain Facial Expressions.</a> </p>