Pupil Cloud

Sometimes we find an offset between the gaze and the action (human error when adjusting the offset on the field) and it would be nice to be able to apply corrections in the cloud, and save them for a given recording.

Currently, full screen playback in Cloud does not occupy the entire screen. True full screen would be preferable for demoing to students/examining recordings more closely

When calculating gaze mapping enrichments, like the Reference Image Mapper or Marker Mapper, there is always a chance that some of the gaze samples are not mapped correctly. Using the available validation tools (e.g. side-by-side view) such errors are easy to spot, but there is no easy way to correct them. Correction can only be done manually in the exported data. There should be a way to correct the possible failure cases: False negatives: a sample has not been mapped at all. False positives: a sample has been mapped when it should not have been. Inaccurate mapping: a sample has been mapped successfully, but its position is a bit off.

Why can't I see the heatmap option in the tools of the enrichment "Reference Image Mapper" in Pupil Cloud? There is only Areas of Interest.

The Reference Image Mapper allows to flexibly map gaze data onto a reference image. In many experimental setups, there are however multiple objects of interest in the reference image (e.g. different elements of a UI, or different items on a shelf). Ultimately, one is interested in relating gaze data to those objects rather than the entire reference image. Thus, a tool that allows to define areas of interest (AOIs) in the reference image, which then maps gaze data onto the individual AOIs would be helpful. This would essentially be an optimized UI for doing what is happening in the "Define AOIs and Calculate Gaze Metrics" How-to Guide. https://docs.pupil-labs.com/invisible/how-tos/advanced-analysis/gaze-metrics-in-aois/

While the automatic tools for gaze mapping in Pupil Cloud work great a lot of the time, they are not applicable in some scenarios: The marker Mapper can not be used if placing markers in the environment is not tolerable. The Reference Image Mapper can not be used in some environments that have very few visual features. An interface for creating a manual gaze mapping would be helpful in such cases. It would be labor-intensive to map manually, but it enables gaze mapping in scenarios that would otherwise not be possible.

Sometimes you may have a recording where unfortunately you recorded something sensible and you would like to remove the scene video from that part, so you do not have to remove the whole rec. A function allowing you remove sections of the scene camera would be nice.

Accidentally uploading a recording to the wrong workspace can easily happen in practice. There is currently no way to transfer recordings between workspaces, so it is difficult to recover such recordings for analysis in Pupil Cloud. The ability to transfer recordings between workspaces to correct such mistakes would be desirable.

When generating the gaze overlay there is an option to obtain the undistorted video, please provide the corrected/undistorted gaze position on a CSV file for download.

Many studies are interested in calculating a standard set of metrics on different subsets of data to compare behavior/performance across those different conditions. A tool that allows to... 1) flexibly split data into subsets based on various categories like splitting wearers into multiple groups splitting up wearers based on metadata (e.g. age) collected via templates. 2) calculate standard gaze metrics on every subset and visualizes the results in a table, bar chart, or similar. Standard metrics could include Total dwell time (total time a subject has been gazing onto an AOI). Time to first fixation (time spend before the subject looked at the AOI for the first time). Average fixation duration (average duration of fixations of a subject on the AOI). Reach (percentage of subjects that have looked at the AOI at least once). ...would be helpful in automating this process.