Emphasizing 3D Properties in Recurrent Multi-View Aggregation for 3D Shape Retrieval
Authors: Cheng Xu, Biao Leng, Cheng Zhang, Xiaochen Zhou
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate ERFA-Net on two popular 3D shape datasets, Model Net and Shape Net Core55, and ERFA-Net outperforms the state-of-the-art methods significantly. Extensive experiments show the effectiveness and robustness of the proposed 3D representation. |
| Researcher Affiliation | Academia | Cheng Xu, Biao Leng, Cheng Zhang, Xiaochen Zhou School of Computer Science and Engineering, Beihang University, Beijing 100191, China {cxu, lengbiao, zcheng, zhouxiaochen}@buaa.edu.cn |
| Pseudocode | No | The paper includes mathematical equations for LSTM but does not provide any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not contain any explicit statement or link regarding the public release of its source code. |
| Open Datasets | Yes | We evaluate ERFA-Net on two popular 3D shape datasets, Model Net and Shape Net Core55... The dataset from SHape REtrieval Contest (SHREC) 2016 is a large-scale 3D shape retrieval track. |
| Dataset Splits | Yes | Model Net Dataset... We adopt the same training and testing split mentioned in (Wu et al. 2015). We randomly select 100 unique shapes per category from the subset, where the first 80 shapes are used for training and the rest for testing... SHREC) 2016... we adopt the official training and testing split method, where the database is split into three parts, 70% shapes used for training, 10% shapes for validation data and the rest 20% for testing. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for experiments (e.g., GPU models, CPU types, memory). |
| Software Dependencies | No | The paper mentions using "caffe (Jia et al. 2014) toolbox to implement and train the network" but does not specify its version number or any other software dependencies with their versions. |
| Experiment Setup | Yes | Each 3D shape is rendered to yield views of 256 × 256 pixels... The length of multi-view image sequence is 12... the input image is resized to 224 × 224... The initial learning rate is 1e-4, which is annealed by 0.5... The weight decay is set to 5e-4 and the momentum is set to 0.9. |