PhysLoc

Physics-violation video clips with spatio-temporal annotations: every invalid clip ships where the violation is, when it happens, and how badly -- all derived from the simulator rather than annotated by hand.

Clips come in twins. A valid clip and its invalid partner share a scene, a seed, and a bit-identical prefix up to t_event; only after that do they differ. That is what makes the difference between them attributable to the intervention and nothing else.

What is here

clips	179 (165 invalid, 14 valid)
pairs	14
scenarios	13
violation families	23
tier	debug
moving camera	31 clips (17%)

Files

• shards/<split>-*.tar -- RGB video and every annotation, one set per split. This is the data. The hub previews index rather than these, so browse the table to find what you want and stream the shard to get it.
• The raw geometry passes (depth, optical flow, normals, object coordinates) are not in this release. They are about 86% of the bytes and are packaged only on request.
• index.parquet -- one row per clip: uids, scenario, family, severity, windows, camera motion, actor shape and material.
• splits/*.txt -- clip uids per split.

Splits leak nothing

Splits are grouped by pair_uid, never by clip. A valid twin and its invalid siblings share every frame before t_event, so splitting them apart would put the answer on the other side.

Pairs are ordered by a hash of their uid and cut at the quantiles WITHIN each scenario, so every split sees every scenario and the proportions are exact. There is no rng, so reproducing this release reproduces its splits. Adding clips moves the boundaries.

Every split ships every annotation. The split is a label, not a filter: you can re-cut it, and you can score any clip in the release. If you need a genuinely blind held-out set -- for a leaderboard others submit to -- strip the annotations from splits/held_out.txt at that point; nothing here has to be regenerated to do it.

Reading a clip

import webdataset as wds

ds = wds.WebDataset("shards/core-000.tar")
for sample in ds:
    meta = sample["meta.json"]      # bytes -> json.loads
    video = sample["rgb.mp4"]
    mask = sample["violation_mask.npz"]

The annotations, and what they are not

• violation_mask -- where the violation is, unioned over BOTH twins. A vanished body has no pixels in the invalid render, which is exactly why the union is needed.
• mask_invalid -- the same footprint, invalid side only.
• severity_map -- continuous, how badly, invalid side only.
• causal_mask -- level 1 the culprit, level 2 a body it disturbed.
• reference_mask -- where the culprit lawfully should have been.
• divergence_map -- |valid - invalid| in pixels. Ships for inspection, not for training: it diverges everywhere downstream of the event, so a model trained on it learns to find the edit rather than the physics.

Scenarios

• barrier_pass (15 clips)
• collision (17 clips)
• drop (15 clips)
• occluder_pass (13 clips)
• pendulum_swing (9 clips)
• pour (18 clips)
• pyramid_impact (14 clips)
• ramp_slide (12 clips)
• resting_table (12 clips)
• rolling_ramp (16 clips)
• shadow_track (11 clips)
• stack_topple (14 clips)
• toss (13 clips)

Violation families

• angular_momentum (4 clips)
• antigravity (6 clips)
• colour_shift (13 clips)
• continuity (13 clips)
• deformation (13 clips)
• dissolve (13 clips)
• fission (7 clips)
• friction (6 clips)
• fusion (2 clips)
• global_gravity (5 clips)
• immutability (13 clips)
• newton1_inertia (6 clips)
• newton2_mass (2 clips)
• non_parabolic (6 clips)
• permanence (13 clips)
• phantom_impulse (15 clips)
• shadow (1 clips)
• shadow_inverted (1 clips)
• shadow_shape (2 clips)
• solidity (9 clips)
• superelastic (6 clips)
• support (17 clips)
• time_slip (6 clips)