3D WM Atomic — v3 (batches b07–b14) Synthetic CAD construction videos paired with per frame atomic action ops , for training video→action (inverse dynamics) and image→video world models. v3 extends the v2 dataset (batches b01–b06) with b07–b14 . Identical format and contract. Format: v2 continuous mm Raw float millimeter op parameters (no quantization); each clip is translation canonicalized so the final mesh bbox is centered at the world origin. The full schema — atomic op vocabulary, scale/camera metadata, the SKETCH BEGIN depth DOF caveat, and the recommended near blank reader filter — is documented in the companion dataset spec.md . Layout Each clip = one (case, view) : 0000.png … NNNN.png — per op construction animation frames (256×256) meta.json — scale / camera / canonicalization + per frame op spans ops.json — atomic op tuples (raw mm floats) + human readable ops human 5 views per case (iso / front / side / top / random) sharing one auto fit scale anchor — identical world diag mm / approx mm per pixel across siblings, only the camera angle differs. Quality gates (at generation time) Geometry validity — OCC BRepCheck ; invalid / self intersecting solids are dropped. Near bla…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy