UCBProject / ProcessedData Phase 1A outputs of the Affordance2Grasp pipeline — used by downstream phases: Phase 2 (training the main method's PointNet++ contact prediction net) reads training fp/{dataset}/{object}.hdf5 . Phase 3 (inference, grasp sampling, sim execution, Sim2Real deployment) reads per object affordance priors from human prior fp/{object}.hdf5 . The upstream object meshes (SAM3D reconstructions) live in a sibling repo, UCBProject/ObjMesh . Layout HDF5 schema training fp/{ds}/{obj}.hdf5 (training ready) key shape dtype meaning point cloud (4096, 3) float32 4096 surface samples of the SAM3D object mesh (metric m, mesh canonical frame) normals (4096, 3) float32 per point unit normals human prior (4096,) float32 per point contact probability in [0, 1] (Gaussian smoothed, per object max normalised) robot gt (4096,) float32 all zero placeholder (no robot ground truth for this regime) force center (3,) float32 centroid of mesh verts with contact smooth = 80th percentile attr object str — object id (e.g. ycb dex 14 or A01001 ) human prior fp/{obj}.hdf5 (inference ready, indexed by object only) Same five fields as above. Same content as training fp/{ds}/{obj}.hdf5 , but flat…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy