p01rand (dense MSE only) under-predicts contact — val area 0.29x GT. contactw1 adds a strong contact-region reconstruction loss (masked MSE, w=1.0, threshold 0.20, no false-contact term) then recovers contact. Both are full 3-view + 2-tactile models.
| run | contact IoU (val) | pred/GT area (val) | pred intensity in contact |
|---|---|---|---|
| p01rand | 0.173 | 0.29 | 5.84 |
| contactw1 | 0.202 | 0.85 | 7.04 |
Columns: GT tactile RGB · GT contact · p01rand pred · contactw1 pred (red = above tau=8 contact threshold).
Matched val windows. Click the play button or scrub; the view clip is the clock, tactile clips follow.
2026-05-10__episode_005.pt · t=604 — GT ‖ prediction. top = middle view, then tactile left / right.
2026-05-10__episode_005.pt · t=606 — GT ‖ prediction. top = middle view, then tactile left / right.
2026-05-10__episode_006.pt · t=1116 — GT ‖ prediction. top = middle view, then tactile left / right.
2026-05-10__episode_006.pt · t=1226 — GT ‖ prediction. top = middle view, then tactile left / right.
2026-05-10__episode_006.pt · t=1228 — GT ‖ prediction. top = middle view, then tactile left / right.
2026-05-10__episode_006.pt · t=1238 — GT ‖ prediction. top = middle view, then tactile left / right.
2026-05-10__episode_006.pt · t=1420 — GT ‖ prediction. top = middle view, then tactile left / right.
2026-05-10__episode_006.pt · t=2460 — GT ‖ prediction. top = middle view, then tactile left / right.
2026-05-10__episode_006.pt · t=2496 — GT ‖ prediction. top = middle view, then tactile left / right.
2026-05-10__episode_006.pt · t=2502 — GT ‖ prediction. top = middle view, then tactile left / right.
Full 40-second autoregressive rollout on held-out episodes — tests whether the recovered contact prediction stays stable over long horizons (it does; no collapse/drift).
Full 40-second autoregressive rollout. top = middle view, then tactile left / right.
Full 40-second autoregressive rollout. top = middle view, then tactile left / right.