---
pretty_name: FoundationPose Paired Synthetic Renders
configs:
- config_name: gso-scenes
  data_files:
  - split: train
    path: gso/scenes/train-*.parquet
  - split: validation
    path: gso/scenes/validation-*.parquet
- config_name: gso-assets
  data_files:
  - split: train
    path: gso/assets/train-*.parquet
- config_name: objaverse-scenes
  data_files:
  - split: train
    path: objaverse/scenes/train-*.parquet
  - split: validation
    path: objaverse/scenes/validation-*.parquet
- config_name: objaverse-assets
  data_files:
  - split: train
    path: objaverse/assets/train-*.parquet
---

# FoundationPose paired synthetic renders

Each scene row contains two synchronized views and per-object pose, mask ID,
bounding box, visibility, and occlusion annotations. Depth is stored as the
original float32 NPY bytes; RGB and uint32 instance masks are stored as PNG
bytes. Matrices use row-major flattened arrays and canonical column-vector
names (`camera_from_world` and `world_from_object`). Raw source transforms are
retained alongside them.

The assets tables contain stable source identifiers only. The render archives
reference external USD geometry and do not include reusable mesh assets.