Annotations for 3D assets used in LychSim, a controllable and interactive simulation framework for vision research. The dataset provides semantic category, canonical scale, pose alignment, and mesh offset for each asset. It was created by author wufeim and was last updated on May 14, 2026.
Use Cases
- Training vision models on semantically annotated 3D objects based on the provided category labels.
- Calibrating object placement and orientation in simulated scenes based on the pose alignment and mesh offset data.
- Developing scene generation tools that respect canonical object scale and front-facing orientation.
Strengths
- Annotations include pose alignment with a defined calibration yaw for consistent object orientation.
- Provides a precomputed mesh offset for precise object spawning relative to a visual bounding box.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
Provenance
- Source
- huggingface
- Collection Method
- Annotations created for assets used in the LychSim framework.
- Freshness
- Last updated 2026-05-14 05:44:04; freshness should be verified.