Learning Part Generation and Assembly for Structure-Aware Shape Synthesis

Authors: Jun Li, Chengjie Niu, Kai Xu11362-11369

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

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate through both qualitative and quantitative evaluations that PAGENet generates 3D shapes with plausible, diverse and detailed structure
Researcher Affiliation Academia Jun Li, Chengjie Niu, Kai Xu* National University of Defense Technology
Pseudocode No The paper describes the network architecture and training details in text and figures, but does not provide structured pseudocode or algorithm blocks.
Open Source Code No The paper does not explicitly state that source code for the methodology is available or provide a link to a code repository.
Open Datasets Yes We train and test our model on the Shape Net part dataset (Yi et al. 2016), which is a subset of the Shape Net dataset (Chang et al. 2015) and provides consistent alignment and semantic labeling for all shapes.
Dataset Splits Yes The dataset is divided into two parts, according to the official training/test split.
Hardware Specification No The paper mentions training times ('average training time is 12 hours for each part generator and 7 hours for part assembler') but does not specify any hardware details like GPU or CPU models.
Software Dependencies No The paper mentions using ADAM for optimization and WGAN-GP for adversarial training, but does not provide specific version numbers for any software, libraries, or frameworks used.
Experiment Setup Yes For all modules, we use ADAM (β = 0.5) for network optimization with an initial learning rate of 0.001. Batch size is set to 32. The parameters in the loss in Equation (1) are set as α1 = 2 and α2 = 1 10 3 for all experiments. The λ in Equation (2) is set to 10 as in (Gulrajani et al. 2017).