Mining and Transferring Feature-Geometry Coherence for Unsupervised Point Cloud Registration

Authors: KeZheng Xiong, Haoen Xiang, Qingshan Xu, Chenglu Wen, Siqi Shen, Jonathan Jun LI, Cheng Wang

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on KITTI and nu Scenes datasets demonstrate that our INTEGER achieves competitive performance in terms of accuracy and generalizability.
Researcher Affiliation Academia a Fujian Key Laboratory of Sensing and Computing for Smart Cities, Xiamen University, China. b Key Laboratory of Multimedia Trusted Perception and Efficient Computing, Ministry of Education of China, Xiamen University, China. c Nanyang Technological University, Singapore. d University of Waterloo, Waterloo, Canada
Pseudocode Yes Algorithm 1: Feature-Geometry Clustering
Open Source Code Yes [Code Release]
Open Datasets Yes We mainly evaluate INTEGER on two challenging public datasets: KITTI[6] and nu Scenes[7].
Dataset Splits Yes Both datasets adhere to official splits.
Hardware Specification Yes The training process takes approximately 6 days on a single NVIDIA RTX 3090 GPU running at 1.70 GHz with 24 Gi B of GPU memory.
Software Dependencies No The paper mentions using FCGF, SC2-PCR, and the sklearn library for t-SNE and KDE, but does not provide specific version numbers for these software components.
Experiment Setup Yes To train INTEGER, we use the SGD optimizer with an initial learning rate of 0.3 and a weight decay of 1e 4. We train INTEGER for 400 epochs with a batch size of 8.