Research map / Point clouds / Point cloud classification
Autoregressive Uncertainty Modeling for 3D Bounding Box Prediction
YuXuan Liu, Nikhil Mishra, Maximilian Sieb, Yide Shentu, Pieter Abbeel, Xi Chen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| predict box parameters autoregressively instead of independent discrete bins (beam search) (ablation) | Discrete box head · COB-3D | IoU: 0.5232 -> 0.6296 | Improved |
| predict box parameters autoregressively instead of independent discrete bins (beam search) (ablation) | Discrete box head · COB-3D | err_quat: 0.2926 -> 0.1598 | Improved |
| use autoregressive beam search instead of Gaussian output | Box head · COB-3D | IoU: 0.3169 -> 0.6296 | Improved |
| use lower quantile (0.1) instead of beam search | Autoregressive FCAF3D · SUN-RGBD | IoG AP_all: 47.17 -> 60.08 | Improved |
| use lower quantile (0.1) instead of beam search | Autoregressive FCAF3D · SUN-RGBD | IoU AP_all: 38.15 -> 31.06 | Hurt |
| use lower quantile (0.1) instead of beam search | Autoregressive PV-RCNN · KITTI hard (Car) | IoU AP: 82.37 -> 59.75 | Hurt |
| use lower quantile (0.1) instead of beam search | Autoregressive PV-RCNN · KITTI hard (Car) | IoG AP: 91.84 -> 96.02 | Improved |
| use quantile 0.1 instead of beam search | Box estimator · COB-3D | IoG: 0.7877 -> 0.9723 | Improved |
| use quantile 0.1 instead of beam search | Box estimator · COB-3D | IoU: 0.6296 -> 0.3821 | Hurt |
| use quantile 0.5 instead of beam search | Box estimator · COB-3D | IoG: 0.7877 -> 0.8295 | Improved |
| add autoregressive head with beam search to FCAF3D | FCAF3D · SUN-RGBD | IoU AP_all: 37.42 -> 38.15 | Improved |
| add autoregressive head with beam search to FCAF3D | FCAF3D · Scannet | IoU AP_all: 43.32 -> 43.85 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- FCAF3D — FCAF3D: Fully Convolutional Anchor-Free 3D Object Detection (ECCV 2022) · better on 3 of 3 shared results
- PVRCNN — PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection (CVPR 2020) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- PointNSP: Autoregressive 3D Point Cloud Generation with Next-Scale Level-of-Detail Prediction (CVPR 2026 · similarity 0.86)
- Adaptive Bounding Box Uncertainties via Two-Step Conformal Prediction (ECCV 2024 · similarity 0.85)
- Variational Multi-scale Representation for Estimating Uncertainty in 3D Gaussian Splatting (NeurIPS 2024 · similarity 0.85)
- SAR3D: Autoregressive 3D Object Generation and Understanding via Multi-scale 3D VQVAE (CVPR 2025 · similarity 0.84)
- An Empirical Study of the Generalization Ability of Lidar 3D Object Detectors to Unseen Domains (CVPR 2024 · similarity 0.84)
- OpenBox: Annotate Any Bounding Boxes in 3D (NeurIPS 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
