Gallery

The frames the statistic actually consumes — everything else is a distance between averages of images like these.

early vs late no-contact frames — **The durability run.** Early frame, the same sensor after 25 h of gripping, and their amplified difference. Diffuse rather than localised — consistent with a remount, not a torn pad. source: panda_durability_test_data/ted_gel

<b>Every gel in the study.</b> Top row markered, bottom row markerless. The dist — **Every gel in the study.** Top row markered, bottom row markerless. The distinction is asserted on every page; this is what it looks like. Note how far apart the datasets sit in colour even within a row, before anything is pressed into the pad. FEATS, durability, cnc_mini_26, FeelAnyForce, Sparsh · 41 kB

<b>Why the gel axis was corrected.</b> One unit, its six gel indices, cycled. In — **Why the gel axis was corrected.** One unit, its six gel indices, cycled. Indices 0-4 carry the dot lattice; index 5 does not, so it is a different gel *model* rather than another pad. Treating all six alike inflated the gel axis roughly twofold. FEATS v24_labels_24_32 · 6 frames · WebP 30 kB, GIF fallback 156 kB

<b>The unit axis.</b> Five units at a fixed gel index, cycled. The differences a — **The unit axis.** Five units at a fixed gel index, cycled. The differences are global and smooth, illumination and white balance, which is why this axis is far more low-frequency than the gel axis. FEATS v24_labels_24_32 · 5 frames · WebP 30 kB, GIF fallback 143 kB

Terms used on this site

out-of-distribution (OOD)	Input that differs from what the estimator was fitted on — here: a different sensor, a different pad, or a different dataset.
reference frame	An image of the gel with nothing pressing it. Reconstruction works on the difference between a frame and this reference, so the reference carries the sensor's own illumination.
markered / markerless gel	GelSight pads either carry a printed lattice of black dots (used to read shear) or do not. It changes the image far more than swapping one pad for another of the same kind.
rms / rms_lowfreq	Root-mean-square pixel difference between two mean images, in 0–255 units. `rms_lowfreq` keeps only the smooth part — the only version that survives comparing images captured at different resolutions.
noise floor	What the statistic reads when nothing has changed: two averages of the same sensor, same pad. Any effect smaller than this is not measurable.
Spearman rho	Rank correlation, −1…1. It asks whether predictions ORDER the true forces correctly, ignoring scale — so it survives datasets with different force units.
in-domain vs transfer	In-domain: fit and score on the same population. Transfer: fit on one, score on another. The gap between them is the cost of the shift.
permutation null	The score the same procedure produces on shuffled data. Without it, a positive correlation can look meaningful when the procedure would produce one anyway.
Mantel permutation	A significance test for distance-vs-distance comparisons, where observations share populations and so are not independent. It relabels the populations instead of the observations.
LUT vs calibration-free	Two ways to turn an image into surface depth. The LUT (lookup table) is calibrated per sensor from known presses; the calibration-free solve is not, which is why the two degrade differently when the sensor changes.
degenerate transfer	The fitted map sends every target sample to one value, so rank correlation is undefined — total failure, not a missing measurement.