DIMS Dashboard — ORTHO

Parent–child dyads play a collaborative tabletop game. This dashboard puts the recording beside the ball's kinematics, the pair's gaze, and the annotations — and quantifies how those signals relate over time.

Twelve game sessions across three dyads.

Run it

python serve.py            # then open http://localhost:8000
python serve.py 8080       # if that port is taken

Nothing to install: the dashboard is plain files and serve.py uses only the standard library. Videos are not in the repository (see Data); every other tab works without them.

What you get

Pick a session from the dropdown, then a point on the timeline — every tab narrows to a window around it.

tab	what it draws
Time series	ball speed and velocity components, and each partner's gaze, on a shared time axis
Trajectory	where the ball went, drawn on the board itself, with the path so far up to the selected moment
RQA	a recurrence plot per signal — when the ball returns to a state it was in before — with recurrence rate, determinism and laminarity tracked over time. The gaze channels get a categorical RQA instead: recurrence means both partners looking at the same region
Cross-Wavelet	coherence between two signals by time and timescale, with a chance level from a Monte Carlo null. A thick line is not evidence on its own — the tab says what share of cells beat chance
ELAN	the annotation tiers, aligned to the video

Three camera perspectives are available per session — wide, parent, child — switchable above the video.

Data

in the repository	not in the repository
time series, gaze, ELAN, transcripts, board images, all analysis output	the video recordings

The participant data here is published with consent. The recordings are not — dims-case.json lists assets/videos as restricted and .gitignore excludes video files, so they cannot be committed by accident.

To watch the video alongside the data, put the files in assets/videos/ locally as {videoID}_{perspective}.mp4, or point data.local.json at wherever you keep them (copy data.local.json.example). Nothing needs to be copied into the repository.

Two sessions — CUF00_L1_A1 and CUF00_L1_A2 — have time series, RQA and transcripts but no ELAN or cross-wavelet yet. Their tabs render empty until the analyses are re-run.

Rebuild the analyses

pip install -e /path/to/dims          # the analyses; not on PyPI yet
python build_assets.py --check        # what would run
python build_assets.py                # run it

That runs RQA and cross-wavelet for what config.json enables, then this study's own categorical gaze RQA from opt/. Order matters between the two and build_assets.py handles it — see opt/README.md.

Rebuild the time series

Only if you are regenerating from the game's own records rather than from the CSVs already here. It needs ortho.db, which this repository does not ship. See tools/README.md.

Layout

config.json          which sessions exist, which analyses run, trajectory geometry
index.html           the page; loads the vendored core, then tabs/
serve.py             local server (stdlib only)
build_assets.py      rebuild the analyses
vendor/              pinned copy of the shared core — never edit, see below
tabs/trajectory.js   this study's own tab
opt/                 this study's own analysis (categorical gaze RQA)
tools/               rebuild the time series from the game database
assets/              data in, analysis out

vendor/ is a pinned copy of dims, verified against its release in CI. A fix belongs upstream and arrives here as a version bump (dims-case sync), never as an edit here.

Configuration

config.json is validated in CI against the shared schema. The keys specific to this study:

key	what it does
perspectives, videoSrcTemplate	the three camera angles and how their filenames are built
include_trajectory, trajectory_settings, trajectory_tracks	the board background and per-session path overrides
include_RQA	the kinematic channels only — the gaze channels are the categorical step's, and listing them here would make both analyses claim the same data