ShapeFlow: Learnable Deformation Flows Among 3D Shapes
Authors: Chiyu Jiang, Jingwei Huang, Andrea Tagliasacchi, Leonidas J. Guibas
NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | As a first experiment, we learn the deformation space for entire classes of shapes from Shape Net [4], and illustrate two downstream applications for such a deformation space: shape generation by deformation, and shape canonicalization. Specifically, we experiment on three representative shape categories in Shape Net: chair, airplane and car. For each category, we follow the official train/test/validation split for the data. We preprocess the geometries into watertight manifolds using the preprocessing pipeline in [38], and further simplify the meshes to 1/10th of the original number of vertices using [63]. The deformation space is learned by deforming random pairs of objects using a hub-and-spokes deformation approach (as described in Section 3.2). More training details for learning the deformation space can be found in Section B.2 (supplementary material). We benchmark against various state-of-the-art shape generation models that outputs voxel grids (3D-R2N2 [5]), upsampled point sets (PSGN [58]), mesh surfaces (DMC [62]) and implicit surfaces (Occ Flow [38]); see quantitative results in Table 1. |
| Researcher Affiliation | Collaboration | Chiyu Max" Jiang UC Berkeley chiyu.jiang@berkeley.edu Jingwei Huang Stanford University jingweih@stanford.edu Andrea Tagliasacchi Google Brain taglia@google.com Leonidas Guibas Stanford University guibas@stanford.edu |
| Pseudocode | No | The paper includes mathematical formulations and descriptions of processes, but it does not feature a dedicated, structured pseudocode or algorithm block labeled as such. |
| Open Source Code | No | The paper does not provide an explicit statement about releasing its source code, nor does it include a link to a repository containing the code for the described methodology. While it cites [63] for a tool used, this is not their own project code. |
| Open Datasets | Yes | As a first experiment, we learn the deformation space for entire classes of shapes from Shape Net [4] |
| Dataset Splits | Yes | For each category, we follow the official train/test/validation split for the data. |
| Hardware Specification | No | The paper does not specify any details regarding the hardware (e.g., CPU, GPU models, memory) used to conduct the experiments. |
| Software Dependencies | No | The paper mentions using 'a modified version of IM-NET [11] as the backbone flow model' and that meshes were simplified 'using [63]'. However, it does not provide specific version numbers for these or any other software dependencies, making it difficult to precisely replicate the environment. |
| Experiment Setup | No | The paper states 'More training details for learning the deformation space can be found in Section B.2 (supplementary material)'. This indicates that specific setup details like hyperparameters are not provided in the main text. |