1 rendering.candidate_depth_renderer

rendering.candidate_depth_renderer

Render valid candidate poses directly from dataset snippets.

CandidateDepthRenderer consumes an EfmSnippetView with an attached mesh and a CandidateSamplingResult. It renders the compact valid candidate table but returns candidate_indices into the full sampled shell so labels and heavy diagnostics can be joined back to candidate provenance, validity masks, and reason codes.

Depth output is metric PyTorch3D z-depth in the physical camera frame. The renderer is used for oracle/evaluation labels; callers decide whether rendered depths or backprojected point clouds are retained in a rollout store.

1.1 Classes

Name Description
CandidateDepths Typed result for candidate depth rendering.
CandidateDepthRenderer High-level wrapper that renders depth for compact valid candidate poses.