Submitted by Manuel Dahnert.
Submission data
| Full name | Li et al. 2015 + RANSAC |
| Description | We use the local feature descriptor from Li et al. 2015 to compute correspondences between Haris keypoints of the CADs and of the scan.
We filter the correspondences and compute a transformation using RANSAC. Since we do not consider real time, we replace the 1-point RANSAC from the paper with a more robust 3-point RANSAC. |
| Input Data Types | Uses Geometry Uses 3D |
| Programming language(s) | C++, PCL |
| Hardware | Intel Xeon E5-1620 v3 |
| Submission creation date | 23 Jan, 2019 |
| Last edited | 24 Jan, 2019 |
CAD Alignment
| Info | avg (per CAD) | avg (per class) | bathtub | bookshelf | cabinet | chair | display | other | sofa | table | trashbin |
| 0.046 | 0.030 | 0.000 | 0.018 | 0.023 | 0.111 | 0.000 | 0.028 | 0.064 | 0.027 | 0.000 |