Gaze3P
Dataset contents: Gaze3P includes gaze data (i.e. where, when, and how a person looks) of 100 participants viewing 1,000 natural images showing different objects, some with private attributes (e.g. credit cards and medical history). The dataset also provides user ratings of perceived privacy on a scale from 1 (very private) to 7 (very safe) for each image.
Due to the lack of available datasets, we present Gaze3P. Our large-scale data collection is essential for deriving statistically significant insights into users’ perceptions of privacy and for empirically validating the use of gaze as a reliable indicator of perceived privacy. By offering a standardized dataset, Gaze3P aims to facilitate reproducible research and enable the development and evaluation of models that infer perceived privacy dynamically and without explicit feedback through gaze patterns.
The full dataset can be downloaded at: https://www.dropbox.com/scl/fo/ijikoxc2yx9ko5gnjy8q2/AJs0YJwuipWyI9FRqBylCIA?rlkey=5vgo5hle0h85sukg9rpvqt5c0&st=wjwn26cy&dl=0.
The data is only to be used for non-commercial scientific purposes. If you use this dataset in a scientific publication, please cite the following paper:
