5 Hits in 2.7 sec

PAConv: Position Adaptive Convolution with Dynamic Kernel Assembling on Point Clouds [article]

Mutian Xu, Runyu Ding, Hengshuang Zhao, Xiaojuan Qi
2021 arXiv   pre-print
We introduce Position Adaptive Convolution (PAConv), a generic convolution operation for 3D point cloud processing.  ...  The key of PAConv is to construct the convolution kernel by dynamically assembling basic weight matrices stored in Weight Bank, where the coefficients of these weight matrices are self-adaptively learned  ...  We have presented PAConv, a position adaptive convolution operator with dynamic kernel assembling for point cloud processing.  ... 
arXiv:2103.14635v2 fatcat:r4mealnimve3rdck5silqiv6w4

Adaptive Channel Encoding for Point Cloud Analysis [article]

Guoquan Xu, Hezhi Cao, Yifan Zhang, Jianwei Wan, Ke Xu, Yanxin Ma
2021 arXiv   pre-print
Attention mechanism plays a more and more important role in point cloud analysis and channel attention is one of the hotspots.  ...  Specifically, a channel-wise convolution (Channel-Conv) is proposed to adaptively learn the relationship between coordinates and features, so as to encode the channel.  ...  Qi, “Paconv: positionAdaptive graph convolution for point cloud analysis,” in adaptive convolution with dynamic kernel assembling on Proceedings of the IEEE International  ... 
arXiv:2112.02509v1 fatcat:d4ou6pd6ybcb5eycmsdi5jvnre

PnP-3D: A Plug-and-Play for 3D Point Clouds [article]

Shi Qiu, Saeed Anwar, Nick Barnes
2021 arXiv   pre-print
With the help of the deep learning paradigm, many point cloud networks have been invented for visual analysis.  ...  In addition to achieving state-of-the-art results on four widely used point cloud benchmarks, we present comprehensive ablation studies and visualizations to demonstrate our approach's advantages.  ...  Cui, “Pointasnl: Robust convolution with dynamic kernel assembling on point clouds,” in point clouds processing using nonlocal neural networks with CVPR, 2021, pp. 3173–3182  ... 
arXiv:2108.07378v2 fatcat:atpotz75ujg3rbcgm3qhkd72ay

TO-Scene: A Large-scale Dataset for Understanding 3D Tabletop Scenes [article]

Mutian Xu, Pei Chen, Haolin Liu, Xiaoguang Han
2022 arXiv   pre-print
To remedy this defect, we introduce TO-Scene, a large-scale dataset focusing on tabletop scenes, which contains 20,740 scenes with three variants.  ...  Experiments show that the algorithms trained on TO-Scene indeed work on the realistic test data, and our proposed tabletop-aware learning strategy greatly improves the state-of-the-art results on both  ...  Xu, M., Ding, R., Zhao, H., Qi, X.: Paconv: Position adaptive convolution with dynamic kernel assembling on point clouds. In: CVPR (2021) 1, 2, 3, 5.2, C.1 56.  ... 
arXiv:2203.09440v2 fatcat:qfuspmnmp5hufnhlm2uuroqeva

Do We Need Anisotropic Graph Neural Networks? [article]

Shyam A. Tailor, Felix L. Opolka, Pietro Liò, Nicholas D. Lane
2022 arXiv   pre-print
models, including the popular GAT or PNA architectures by using spatially-varying adaptive filters.  ...  EGC achieves higher model accuracy, with lower memory consumption and latency, along with characteristics suited to accelerator implementation, while being a drop-in replacement for existing architectures  ...  Paconv: Position adaptive convolu- tion with dynamic kernel assembling on point clouds, 2021.  ... 
arXiv:2104.01481v5 fatcat:hzfr4th5lndxzniw3j6ua3mrlu