Optimizing MediaPipe for the assessment of hand trajectories using a touchscreen shape-tracing task
Bibliographic record
Abstract
MediaPipe is an artificial intelligence-based system that offers a markerless, lightweight approach to motion capture. Yet, its optimal pipeline and accuracy against a known standard for investigations of upper-limb movements is unknown. We aimed to 1) determine optimal post-processing parameters for assessing hand/arm movements via MediaPipe, and 2) evaluate MediaPipe against a known standard. Participants (N = 4) performed one block (10 trials) of a touchscreen-based shape-tracing task. Trials were captured by a camera (GoPro Hero8, 30FPS), with videos cropped to 2sec clips (one trial per video). 1) Two clips were processed via MediaPipe to obtain x-y coordinate data of hand trajectories. The following post-processing parameters were applied: MediaPipe and touchscreen generated coordinates were resampled using spline interpolation (250 datapoints; original reference frames) and normalised. Following Procrustes transformations, root mean squared error (RMSE; primary outcome measure) was calculated for coordinates generated by MediaPipe vs. the touchscreen computer. RMSE decreased with post-processing (RMSEraw=103.3±7.71px, RMSEpost-processed=1.3±0.15px, d>1). 2) We applied our pipeline to 25 clips and conducted an equivalence test between coordinates generated by MediaPipe vs. the touchscreen computer. Preliminary findings indicate accuracy differed between MediaPipe and the touchscreen computer, but the true difference was between 0-2px (t(24) = -31.0, p < .001; mean RMSE=1.2, 90%CI[1.19, 1.27]). This work identifies key post-processing parameters for MediaPipe applied to evaluate upper-limb movements. Future work will quantify the extent to which MediaPipe differs from a known standard across the full dataset, overall informing applications of MediaPipe in investigations of upper-limb movement.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.010 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.005 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".