Learning Multi-resolution Functional Maps with Spectral Attention for Robust Shape Matching
Authors: Lei Li, Nicolas Donati, Maks Ovsjanikov
NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate the superior performance of our approach through experiments on a suite of challenging near-isometric and non-isometric shape matching benchmarks. |
| Researcher Affiliation | Academia | Lei Li LIX, École Polytechnique, IP Paris lli@lix.polytechnique.fr Nicolas Donati LIX, École Polytechnique, IP Paris nicolas.donati@polytechnique.edu Maks Ovsjanikov LIX, École Polytechnique, IP Paris maks@lix.polytechnique.fr |
| Pseudocode | No | The paper describes the pipeline and steps textually and visually with a diagram, but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Our code and data are publicly available1. 1https://github.com/craigleili/Attentive FMaps |
| Open Datasets | Yes | We perform evaluations on widely adopted human shape matching datasets such as FAUST [2] and SCAPE [51] as well as a non-isometric dataset SHREC 19 [52]... Our code and data are publicly available1. 1https://github.com/craigleili/Attentive FMaps |
| Dataset Splits | No | The paper provides training and testing splits (e.g., 'the FAUST dataset... is split into 80 and 20 shapes for training and testing, respectively'), but does not explicitly detail a separate validation split with percentages or sample counts. |
| Hardware Specification | No | The paper states 'More details can be found in the supplementary material' regarding compute resources, but no specific hardware models (e.g., GPU or CPU types) are detailed in the main text. |
| Software Dependencies | No | The paper mentions 'We implement our network with Py Torch [49]' and 'We use ADAM [50] for optimization', but does not provide specific version numbers for PyTorch or any other software dependencies. |
| Experiment Setup | Yes | We use α = 10 3 in Eq. (1). During pre-processing, we perform a one-time eigendecomposition of the Laplacian on each shape and take the first 200 eigenfunctions... We compute n = 20 multi-resolution functional maps with sizes ranging from 10 10 to 200 200 with a step size τ = 10. We set the batch size to 1 and use ADAM [50] for optimization. We use the WKS [29] descriptor as input to the network for all experiments... |