Model
3 pose detection model for retail objects.
Use the NGC CLI to download:
Copied!
| Field | Response |
|---|---|
| Intended Application(s) & Domain(s): | Detects the projections of 3D keypoints, estimates a 6-Degrees of Freedom (DoF) pose |
| Model Type: | Category-Level Object Pose Estimation |
| Intended Users: | Developers working in smart spaces, retail, and industrial applications |
| Output: | 3D Bounding-Box, Cuboid Dimensions and 6 Degrees of Freedom (DoF) |
| Describe how the model works: | Single RGB imagedetects the projections of 3D keypoints, estimates a 6-DoF pose, and regresses the relative 3D bounding cuboid dimensions within a single RGB image. |
| Technical Limitations: | Model should not be used in low-light, low contrast lighting conditions, with warped or motion-induced images, and with very small and very large objects. Model accuracy may vary with occlusion or truncation. |
| Verified to have met prescribed NVIDIA standards: | Yes |
| Performance Metrics: | 3D Intersection-over-Union (IoU), 2D MPE (Mean Pixel Projection Error), Average Precision (AP) Azimuth Error, Average Precision (AP) Elevation Error |
| Potential Known Risks: | This model may fail to detect projections of 3D keypoints. |
| Licensing: | NVIDIA Open Model License |