Polyhedral Complex Derivation from Piecewise Trilinear Networks

Authors: Jin-Hwa Kim

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We empirically validate correctness and parsimony through chamfer distance and efficiency, and angular distance, while examining the correlation between the eikonal loss and the planarity of the hypersurfaces. and 6 Experiment
Researcher Affiliation Collaboration Jin-Hwa Kim NAVER AI Lab & SNU AIIS Republic of Korea j1nhwa.kim@navercorp.com
Pseudocode Yes Algorithm 1 Polyhedral Complex Derivation
Open Source Code Yes The code is available at https://github.com/naver-ai/tropical-nerf.pytorch.
Open Datasets Yes Table 1 and Figure 4 show the results for the Standford 3D Scanning repository [31]. and The Stanford 3D Scanining repository can be freely download via http://graphics.stanford.edu/ data/3Dscanrep/.
Dataset Splits No The paper does not explicitly specify train/validation/test splits for the neural network training.
Hardware Specification Yes We conducted all experiments with an NVIDIA V100 32GB.
Software Dependencies Yes We use the official Python package of tinycudnn 4 for the Hash Grid module.
Experiment Setup Yes We used the number of layers L of 3 and hidden size H of 16 for the networks, and ϵ of 1e-4 for the sign-vectors (ref. Definition 3.4). The weight for the eikonal loss is 1e-2. For Hash Grid, the resolution levels of 4, feature size of 2, base resolution Nmin of 2, and max resolution Nmax of 32 by default.