PIE-NET: Parametric Inference of Point Cloud Edges

Authors: Xiaogang Wang, Yuelang Xu, Kai Xu, Andrea Tagliasacchi, Bin Zhou, Ali Mahdavi-Amiri, Hao Zhang

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We train and evaluate our method on the ABC dataset, the largest publicly available dataset of CAD models, via ablation studies and compare our results to those produced by traditional (non-learning) processing pipelines, as well as a recent deep learning-based edge detector (ECNet).
Researcher Affiliation Collaboration 1State Key Laboratory of Virtual Reality Technology and Systems, Beihang University 2Simon Fraser University 3National University of Defense Technology 4Google Research 5Tsinghua University
Pseudocode No The paper describes the network architecture and mathematical formulations but does not provide any pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any explicit statement about releasing source code or a link to a code repository.
Open Datasets Yes We train PIE-NET on the recently released ABC dataset by [4], which is composed of more than one million feature-rich CAD models with parametric edge representations. The ABC dataset [46], a large-scale CAD mechanical part dataset containing ground-truth annotations for edges and corner points.
Dataset Splits No The paper mentions training and evaluating on the ABC dataset but does not explicitly provide the specific percentages or counts for training, validation, and test splits.
Hardware Specification Yes Training times for point classification,the open and closed curve proposal networks were about 23, 12, and 8 hours, respectively, for 100 epochs, on an NVIDIA TITIAN X GPU.
Software Dependencies No The paper mentions software components and algorithms like Point Net++, focal loss [48], and smooth L1 loss [49, 44] but does not specify their version numbers.
Experiment Setup Yes We set the parameters τe=0.7 and τc=0.9 throughout our experiments. We set wm=1, wc=1, and wp=10 throughout our experiments. We use τγ=0.6 and τiou=0.6 for all our experiments.