Keypoint Data

[pauladkisson]

Notes for Keypoint Tracking Data

[pauladkisson]

Paper Notebook References

See Extended Data Figure 3, Methods: Capturing 3D keypoints, and fig_s03a, s03b, s03c-keypoints-and-dlight.ipynb.

[pauladkisson]

File Description

• video-image.p = Example image of 15 keypoints on 6 IR Cameras
• photometry-dls-dlight-9.p, ..., photometry-dls-dlight-13.p = photometry sessions for 5 mice = dictionary w/ fields
  • positions_median = array of 3D positions for each keypoint during the session (T x 15 x 3)
  • positions_mean = same as above except using the mean positions and not used in analysis/plotting notebook --> ignore
  • positions_medfilter = very similar to other positions variables and not used in analysis/plotting notebook --> ignore
  • dlight = raw photometry signal used to calculate signal_dff (T x 1)
  • uv = raw uv reference signal used to calculate reference_dff (T x 1)

Notebook Notes

Mapping from keypoint index to name:

mapping = {
    0: 'spine',
    1: 'spine',
    2: 'spine',
    3: 'tail base',
    4: 'tail tip',
    5: 'head center',
    6: 'left ear',
    7: 'right ear',
    8: 'noce', # [sic] (probably supposed to be nose)
    9: 'left hind ankle',
    10: 'left hindpaw',
    11: 'right hind ankle',
    12: 'right hindpaw',
    13: 'left forepaw',
    14: 'right forepaw',
}

[pauladkisson]

Conclusions:

• Keypoint positions should be represented in ndx-pose following example from keypoint-moseq
• Photometry is a bit hard to deal with since its too different to really use the FiberPhotometryInterface we already have, but maybe these sessions are duplicated in the dlight_raw_data/*.parquet files?
• Session/Subject metadata is missing --> need to ask for that
• Can use notebook/figures to validate