FINet: Dual Branches Feature Interaction for Partial-to-Partial Point Cloud Registration

Authors: Hao Xu, Nianjin Ye, Guanghui Liu, Bing Zeng, Shuaicheng Liu2848-2856

AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments demonstrate that our method performs higher precision and robustness compared to the state-of-the-art traditional and learning-based methods.
Researcher Affiliation Collaboration 1University of Electronic Science and Technology of China 2Megvii Technology
Pseudocode No The paper describes the methodology using textual descriptions and mathematical formulas, but does not include any explicit pseudocode or algorithm blocks.
Open Source Code Yes Code is available at https://github.com/megvii-research/FINet.
Open Datasets Yes Model Net40. (Wu et al. 2015) includes CAD models from 40 object categories. We use the data from OMNet (Xu et al. 2021), where 8 axisymmetrical categories are removed to avoid the ill-posed problem. 7Scenes. (Shotton et al. 2013) is a widely used benchmark where data is captured by a Kinect camera in 7 indoor scenes.
Dataset Splits Yes We use the official train/test splits, resulting in 4,196 training, 1,002 validation, and 1,146 test objects.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions the use of Adam optimizer but does not specify versions of any software dependencies like programming languages, libraries, or frameworks.
Experiment Setup Yes We run 4 iterations of alignment. Adam optimizer (Kingma and Ba 2015) is used with lr = 10^-4. The batch size is 64, and training for 260k steps. ... The dropout ratio is set to 0.3. ... the factor λ is empirically set to 4.0 in all our experiments. ... the factors β and γ are set to 10^-3.