This repository contains the BubbleView click data associated with our BubbleView TOCHI paper.

If you use any of this data, please consider citing:

@article{kim2017bubbleview,
  title={BubbleView: an interface for crowdsourcing image importance maps and tracking visual attention},
  author={Kim, Nam Wook and Bylinskii, Zoya and Borkin, Michelle A and Gajos, Krzysztof Z and Oliva, Aude and Durand, Fredo and Pfister, Hanspeter},
  journal={ACM Transactions on Computer-Human Interaction (TOCHI)},
  volume={24},
  number={5},
  pages={36},
  year={2017},
  publisher={ACM}
}

We include .m files to demonstrate how to plot the ground truth fixations and collected BubbleView clicks side-by-side for comparison. The comments in these files indicate which experiments from the paper each of the data files correspond to. The names of the files match the dataset and image set (the image sets are in the targets_[datasetname] directories). For instance, run fiwi.m to plot the side-by-side comparisons of ground truth fixation data (stored in fiwi_eyetrack.mat) and BubbleView clicks collected with different parameter settings (fiwi*x*.mat). The fixations and clicks will be plotted on top of the images stored in targets_fiwi.

In the following table we indicate which experiment from the paper corresponds to which files:

Note that addendum corresponds to an extra experiment that was run post-publication, the results of which are written up in bubbleview-addendum.pdf.

In the following table we indicate the corresponding ground truth files that BubbleView clicks were compared against (taken from the original datasets):

For completeness, in the following table we include the BubbleView parameters with which each data file was collected.