One Ring to Rule Them All: Certifiably Robust Geometric Perception with Outliers
Authors: Heng Yang, Luca Carlone
NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Numerical experiments across different perception problems, including single rotation averaging, shape alignment, 3D point cloud and mesh registration, and high-integrity satellite pose estimation, demonstrate the tightness of our relaxations, the correctness of the certification, and the scalability of the proposed dual certifiers to large problems, beyond the reach of current SDP solvers.1 |
| Researcher Affiliation | Academia | Heng Yang and Luca Carlone Laboratory for Information and Decision Systems (LIDS) Massachusetts Institute of Technology {hankyang,lcarlone}@mit.edu |
| Pseudocode | No | The paper describes algorithms and mathematical formulations but does not include any clearly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | 1Code available at https://github.com/MIT-SPARK/Certifiably Robust Perception. |
| Open Datasets | Yes | Satellite Pose Estimation. Satellite pose estimation using monocular vision is a crucial technology for many space operations [86, 28]. We use Shape Alignment (Example 2) to perform 6D pose estimation from satellite images in the SPEED dataset [86] (see Fig. 2(d)). |
| Dataset Splits | No | The paper describes generating inliers and outliers and uses Monte Carlo runs, but does not provide specific percentages or counts for training, validation, and test splits from a larger dataset. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware used to run the experiments, such as CPU or GPU models, or memory specifications. |
| Software Dependencies | No | We model the sparse moment relaxation (4) using YALMIP [67] in Matlab and solve the resulting SDPs using MOSEK [6]. DRS is implemented in Matlab using γτ = 2. The paper mentions software names (YALMIP, Matlab, MOSEK) but does not provide specific version numbers for these dependencies. |
| Experiment Setup | Yes | The threshold in problem (TLS) is set to c = 1 for all applications, and βi, i = 1, . . . , N, is set to be proportional to the inlier noise. The relative weight between point-to-plane distance and normal-tonormal distance in MR is set to wi = 1, i = 1, . . . , N. |