react-force-recovery / results.html
yxma's picture
force recovery: methods, evaluation, debug log
55f7a18 verified
Raw
History Blame
10.3 kB
<!DOCTYPE html>
<html lang="en"><head><meta charset="utf-8">
<meta name="viewport" content="width=device-width,initial-scale=1">
<title>Results β€” Force Estimation on GelSight Mini</title>
<link rel="preconnect" href="https://fonts.googleapis.com">
<link href="https://fonts.googleapis.com/css2?family=Fraunces:opsz,wght@9..144,600;9..144,700&family=IBM+Plex+Mono:wght@400;500&family=IBM+Plex+Sans:wght@400;600&family=Noto+Serif+SC:wght@600;700&family=Noto+Sans+SC:wght@400;500&display=swap" rel="stylesheet">
<style>
:root{--paper:#0d1526;--ink:#dce7f5;--dim:#7c8db0;--grid:#1a2540;
--force:#ffb347;--target:#ff7847;--ok:#7be0a0;--card:#111b31;--line:#24345;}
*{box-sizing:border-box}
body{margin:0;background:var(--paper);color:var(--ink);
font-family:'IBM Plex Sans','Noto Sans SC',sans-serif;line-height:1.65;font-size:16px;
background-image:linear-gradient(var(--grid) 1px,transparent 1px),
linear-gradient(90deg,var(--grid) 1px,transparent 1px);background-size:44px 44px}
.wrap{max-width:880px;margin:0 auto;padding:0 24px 80px}
h1,h2{font-family:'Fraunces','Noto Serif SC',serif}
h1{font-size:2.1rem;font-weight:700;line-height:1.2;margin:0 0 10px}
h2{font-size:1.25rem;font-weight:600;margin:46px 0 8px;color:var(--force)}
.kicker{font-family:'IBM Plex Mono',monospace;font-size:.76rem;letter-spacing:.16em;
text-transform:uppercase;color:var(--force);margin-bottom:14px}
header{padding:60px 0 26px;border-bottom:1px dashed var(--line)}
.sub{color:var(--dim);max-width:720px}
.card{background:var(--card);border:1px solid var(--line);border-radius:10px;
padding:18px 22px;margin:14px 0}
.step{display:grid;grid-template-columns:34px 200px 1fr;gap:12px;
padding:10px 0;border-bottom:1px solid var(--line);align-items:baseline}
.step:last-child{border-bottom:none}
.step .n{font-family:'Fraunces',serif;font-size:1.25rem;color:var(--target)}
.step .t{font-family:'IBM Plex Mono',monospace;font-size:.85rem}
.step .d{color:var(--dim);font-size:.9rem}
table{border-collapse:collapse;width:100%;font-size:.88rem;margin:10px 0;
font-variant-numeric:tabular-nums}
th,td{padding:7px 10px;border-bottom:1px solid var(--line);text-align:right}
th{color:var(--dim);font-family:'IBM Plex Mono',monospace;font-size:.72rem;
text-transform:uppercase;letter-spacing:.08em}
td:first-child,th:first-child{text-align:left}
img,video{max-width:100%;height:auto;border-radius:8px;border:1px solid var(--line);
display:block;margin:12px auto}
a{color:#4fd8e0}a:hover{color:var(--force)}
.pill{display:inline-block;border:1px solid var(--line);border-radius:999px;
padding:3px 12px;font-size:.75rem;font-family:'IBM Plex Mono',monospace;
color:var(--dim);margin-right:8px;margin-top:12px;text-decoration:none}
code{font-family:'IBM Plex Mono',monospace;background:#0a1120;padding:1px 5px;
border-radius:4px;font-size:.85em}
.ok{color:var(--ok)}.warn{color:#e0b34f}
footer{margin-top:60px;padding-top:18px;border-top:1px dashed var(--line);
color:var(--dim);font-size:.8rem}
@media(max-width:640px){.step{grid-template-columns:28px 1fr;grid-auto-flow:dense}
.step .d{grid-column:2}
h1{font-size:1.45rem}
table{display:block;overflow-x:auto;white-space:nowrap;-webkit-overflow-scrolling:touch}
th,td{white-space:normal;min-width:110px}}
.matrix td.best{color:var(--ok);font-weight:700}
.matrix td.dead{color:#c0392b}
</style></head><body><div class="wrap">
<header>
<div class="kicker">React force recovery Β· results</div>
<h1>Three estimators Γ— three ground-truth datasets</h1>
<p class="sub">Every method evaluated on every dataset with force-sensor labels,
predicted vs ground truth per dataset β€” so dataset quality is controlled within
each row, and differences between panels are differences between methods.</p>
<a class="pill" href="method.html">β†– how the method is designed</a>
<a class="pill" href="index.html">overview</a>
<a class="pill" href="gallery.html">gallery</a>
<a class="pill" href="results_zh.html">δΈ­ζ–‡</a>
</header>
<h2>The matrix (Spearman ρ, predicted vs F/T ground truth)</h2>
<div class="card"><table class="matrix">
<tr><th>dataset (gel type)</th><th>Ours β€” physics,<br>0 training frames</th>
<th>FEATS U-net<br>(trained: marker gel)</th>
<th>FeelAnyForce<br>(trained: markerless)</th></tr>
<tr><td>FEATS val (marker)</td><td>0.77</td>
<td class="best">0.96 Β· in-domain</td><td class="dead">0.43</td></tr>
<tr><td>FoTa cnc_Mini (markerless)</td><td>0.94 (in view)</td>
<td class="dead">0.07</td><td class="best">0.83</td></tr>
<tr><td>GlowTact (markerless)</td><td>0.98</td>
<td class="dead">0.04</td><td class="best">0.90</td></tr>
</table>
<p>The pattern is the finding: <b>each network dominates its own gel domain and
collapses outside it</b> (FEATS 0.96 β†’ 0.04–0.07; FeelAnyForce 0.90 β†’ 0.43),
while <b>the physics pipeline is the only estimator that works everywhere</b>
(0.74–0.99) β€” it never sees training data, so it has no domain to leave.</p>
</div>
<h2>FEATS dataset β€” marker-dot gel</h2>
<div class="card"><img src="assets/results_feats.png" alt="FEATS dataset panels">
<p class="footnote">In-domain, the FEATS U-net is excellent (ρ=0.96) β€” the
negative results elsewhere are domain effects, not a weak model. FeelAnyForce,
markerless-trained, degrades on the dotted gel (0.43): the same knife cuts both
ways.</p></div>
<h2>FoTa cnc_Mini β€” markerless gel</h2>
<div class="card"><img src="assets/results_cnc.png" alt="cnc_Mini panels">
<p class="footnote">Hard conditions: only 4 contact-free frames, 62% of presses
near the pad border β€” the press grid is larger than the field of view (see
<a href="debug_pipeline.html">pipeline debug</a>). Strictly in view our ρ is
0.94 (MAE 0.26 N); FeelAnyForce reaches 0.92 non-edge.</p></div>
<h2>GlowTact β€” markerless gel, cleaned</h2>
<div class="card"><img src="assets/results_glowtact.png" alt="GlowTact panels">
<p class="footnote">Friendliest ground truth (centred presses, 10 free frames,
0–20 N). Per-indenter, calibrated within each family under the physical scope
(contact fully in view, gel not bottomed out), ρ is 0.975–0.992 across all six
indenters with MAE ≀ 0.73 N. Caveat: that calibration is refit per family, so
it measures rank agreement within a group, not a transferable absolute-newton
scale β€” see <a href="method.html">per-dataset calibration</a>.</p></div>
<h2>Sparsh (Meta) β€” a fourth dataset, and where the method breaks</h2>
<div class="card"><img src="assets/results_sparsh.png" alt="Sparsh results">
<p class="footnote">10 gel pads (6 sphere, 2 flat, 2 sharp), force in newtons.
Our GlowTact table applied to this foreign sensor reaches &rho;=0.878 on
in-view frames. Rebuilding the table from <b>Sparsh's own sphere presses</b>
β€” 708 frames, radius fitted at R=2.44 mm from a&sup2;=d(2R&minus;d) β€” takes it
to <b>&rho;=0.968, MAE 0.042 N</b>, against a labels-shuffled-within-pad
control of 0.23. Fitting on one pad and applying it <b>unchanged</b> to another
costs nothing measurable (0.96&ndash;0.98 everywhere): one table, six gel pads.</p>
<img src="assets/sparsh_dome.png" alt="dome before and after">
<p class="footnote">Why the table matters more than the fit: with the wrong
sensor's table a sphere press integrates to a <b>bilobed shape with a central
dip</b>; with the self-calibrated table it is a single dome matching the
analytic spherical cap (residual RMS 0.179 &rarr; <b>0.0545 mm</b>). Measured
before any integration, the LUT gradient sits <b>93.3&deg;</b> from the analytic
sphere gradient β€” chance is 90&deg; β€” and self-calibration brings it to
<b>4.5&deg;</b> (within 30&deg;: 15% &rarr; 99%). Reverse control: the Sparsh
table fails on GlowTact frames too, so this is a per-sensor property, not a
bad table.</p>
<p class="footnote"><b>What this costs and where it still fails.</b> The price
is one set of sphere presses with logged depth on the target sensor: this is
<b>calibrate once per sensor</b>, not zero-shot. Rank order transfers across
indenter shapes but <i>absolute newtons do not</i> β€” a sphere-fitted model
applied to a flat punch degrades to MAE 0.37&ndash;0.40 N, and it got
<i>worse</i> with the correct table, because true geometry widens the real
feature-scale gap between a sphere and a punch. The sharp indenter is
unsupported (in-view &rho; 0.58). Shear stays out of reach by construction β€”
the top shear decile keeps 1.6&times; the residual whichever table is used.
Frames are restricted to a visible contact disc: 36% of presses show none and
11% are clipped, and the clipped subset carries the <i>highest</i> median force
yet scores worse, so this is visibility, not force-range filtering. Three
dataset defects had to be fixed first: flat/sharp trajectories ship 5 more
frame indices than forces (silently drifting the labels, &rho;&asymp;0 until
paired within each trajectory), sharp/batch_2 is stored BGR while the other
nine are RGB, and flat/batch_2 ships only 3 of 4 image files.</p></div>
<h2>React β€” no ground truth, so: do independent methods agree?</h2>
<div class="card"><img src="assets/results_react.png" alt="React agreement">
<p class="footnote">On the dataset we actually care about, the two surviving
estimators β€” physics (zero training) and FeelAnyForce (200K frames) β€” agree at
ρ=0.91, and FeelAnyForce reads β‰ˆ0 N on every frame the physics pipeline calls
contact-free. Neither can copy the other's mistakes.</p></div>
<h2>Takeaway</h2>
<div class="card"><p style="margin-top:0"><b>For labelling React</b> (markerless
Mini): FeelAnyForce as the primary labeller, the physics pipeline as an
independent audit, disagreement rows flagged. <b>For any new gel or sensor</b>
where no trained model matches the domain: the physics pipeline is the only
option that works out of the box β€” and its FEATS-dataset score (0.74) shows
what it does on a domain nobody tuned it for.</p></div>
<p><a href="debug_pipeline.html"><b>Pipeline debug page</b></a>: raw
image &rarr; force step by step on all three datasets, and the cnc
field-of-view ablation (in-view &rho;=0.94).</p>
<footer>React force recovery Β· <a href="index.html">overview</a> Β·
<a href="method.html">method design</a> Β· <a href="gallery.html">gallery</a> Β·
data: FEATS (2411.03315) Β· FoTa/T3 (2406.13640) Β· GlowTact
(dacongming666/GlowTact_Datasets) Β· FeelAnyForce (2410.02048)</footer>
</div></body></html>