Tensor-Based Synchronization and the Low-Rankness of the Block Trifocal Tensor
Authors: Daniel Miao, Gilad Lerman, Joe Kileel
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental comparisons with stateof-the-art global synchronization methods on real datasets demonstrate the potential of this algorithm for significantly improving location estimation accuracy. |
| Researcher Affiliation | Academia | School of Mathematics, University of Minnesota (miao0022@umn.edu, lerman@umn.edu) Department of Mathematics and Oden Institute for Computational Engineering and Sciences, University of Texas at Austin (jkileel@math.utexas.edu) |
| Pseudocode | Yes | Algorithm 1 HOSVD-HT |
| Open Source Code | Yes | We include our code in the following github repository: Trifocal Sync. |
| Open Datasets | Yes | We conduct experiments of Algorithm 2 on two benchmark real datasets, the EPFL datasets [42] and the Photo Tourism datasets [11]. |
| Dataset Splits | No | The paper describes the use of 'EPFL datasets' and 'Photo Tourism datasets' and discusses experimental procedures such as feature detection, matching, and synchronization. However, it does not provide explicit details on how these datasets were split into training, validation, and test sets, or if standard predefined splits were used with citations. |
| Hardware Specification | Yes | We test our full pipeline on two EPFL datasets on a personal machine with 2 GHz Intel Core i5 with 4 cores and 16GB of memory. |
| Software Dependencies | No | The paper mentions specific software tools like 'Glue Stick [45]' and 'GC-ransac [49]' but does not provide their specific version numbers or other software dependencies with versioning (e.g., Python, PyTorch, etc.). |
| Experiment Setup | Yes | We apply STE from [41] to find 40% of the correspondences as inliers, then use at most 30 inlier point correspondences to linearly estimate the trifocal tensor. To refine the estimates, we apply bundle adjustment on the inliers and delete triplets with reprojection error larger than 1 pixel. |