SnowflakeNet: Point Cloud Completion by Snowflake Point Deconvolution with Skip-Transformer

Peng, Xiang; Wen, Xin; Liu, Yu‐Shen; Cao, Yan‐Pei; Wan, Pengfei; Zheng, Wen; Han, Zhizhong

doi:10.1109/iccv48922.2021.00545

Cited by 157 publications

(152 citation statements)

References 41 publications

Supporting

Mentioning

150

Contrasting

Order By: Relevance

“…8. Compared with recent GFV based method [40], the distribution of SFM projections is more uniform at instance level across different categories, demonstrating that SFM has learned sufficient local information to distinguish different shapes. Note that we use the same encoder as [40] when conducting the comparison.…”

Section: Ablation Studymentioning

confidence: 85%

“…VRCNet [21] and ASFM-Net [39] improve the global features by narrow the distribution difference between the incomplete and complete point cloud. CRN [32] and SnowflakeNet [40] generate the complete point cloud following in a progressive manner. Other notable work such as PMP-Net [37] formulate the completion as a point cloud deformation process, where point-wise moving paths are predicted to move each point of the incomplete input to complete the point cloud.…”

Section: Related Workmentioning

confidence: 99%

“…Feature structuring network. We compare the performance of point cloud completion to quantitatively validate the effect of the FSNet, where the structured feature map SFM is generated with different numbers of channels h. Additionally, we replace the FSNet with max-pooling operation to produce a global feature vector (GFV) as the global representation instead of SFM, and the coarse point cloud is generated by the module in previous method [40]. As shown in Tab.…”

Section: Ablation Studymentioning

confidence: 99%

“…It shows that as the the dense features are progressively corrected via the feedback mechanism, the local details and surface uniformity of the dense point cloud are gradually improved. Note that only the dense features from the last feedback block are fed to the offset regression module during both the training and test stages, and there is no need to supervise the intermediate results as previous methods do [11,40].…”

Section: Ablation Studymentioning

confidence: 99%

See 3 more Smart Citations

Point cloud completion on structured feature map with feedback network

Zejia¹,

Huang²,

Ma³

et al. 2022

Preprint

View full text Add to dashboard Cite

In this paper, we tackle the challenging problem of point cloud completion from the perspective of feature learning. Our key observation is that to recover the underlying structures as well as surface details given a partial input, a fundamental component is a good feature representation that can capture both global structure and local geometric details. Towards this end, we first propose FSNet, a feature structuring module that can adaptively aggregate point-wise features into a 2D structured feature map by learning multiple latent patterns from local regions. We then integrate FSNet into a coarse-to-fine pipeline for point cloud completion. Specifically, a 2D convolutional neural network is adopted to decode feature maps from FSNet into a coarse and complete point cloud. Next, a point cloud upsampling network is used to generate dense point cloud from the partial input and the coarse intermediate output. To efficiently exploit the local structures and enhance the point distribution uniformity, we propose IFNet, a point upsampling module with self-correction mechanism that can progressively refine details of the generated dense point cloud.We conduct both qualitative and quantitative experiments on ShapeNet, MVP, and KITTI datasets, which demonstrate that our method outperforms state-of-the-art point cloud completion approaches.

show abstract

Section: Ablation Studymentioning

confidence: 85%

Section: Related Workmentioning

confidence: 99%

Section: Ablation Studymentioning

confidence: 99%

Section: Ablation Studymentioning

confidence: 99%

See 2 more Smart Citations

Point cloud completion on structured feature map with feedback network

Zejia¹,

Huang²,

Ma³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Based on this feature, PMP-Net (Wen et al 2020b) completes the entire object gradually from the observed regions to the nearest occluded regions. SnowflakeNet (Xiang et al 2021) also uses the PointNet++ features to split points in the coarsely reconstructed object to execute the completion progressively. In addition, building a similar feature as PointNet, ME-PCN (Gong et al 2021) takes both the occupied and the empty regions on the depth image as input for 3D completion, showing the advantage of masking the empty regions in completion.…”

Section: Point Cloudmentioning

confidence: 99%

SoftPool++: An Encoder–Decoder Network for Point Cloud Completion

et al. 2022

View full text Add to dashboard Cite

We propose a novel convolutional operator for the task of point cloud completion. One striking characteristic of our approach is that, conversely to related work it does not require any max-pooling or voxelization operation. Instead, the proposed operator used to learn the point cloud embedding in the encoder extracts permutation-invariant features from the point cloud via a soft-pooling of feature activations, which are able to preserve fine-grained geometric details. These features are then passed on to a decoder architecture. Due to the compression in the encoder, a typical limitation of this type of architectures is that they tend to lose parts of the input shape structure. We propose to overcome this limitation by using skip connections specifically devised for point clouds, where links between corresponding layers in the encoder and the decoder are established. As part of these connections, we introduce a transformation matrix that projects the features from the encoder to the decoder and vice-versa. The quantitative and qualitative results on the task of object completion from partial scans on the ShapeNet dataset show that incorporating our approach achieves state-of-the-art performance in shape completion both at low and high resolutions.

show abstract

PVT: Point‐voxel transformer for point cloud learning

Zhong-yu

Wan

Shen

et al. 2022

Int J of Intelligent Sys

View full text Add to dashboard Cite

The recently developed pure transformer architectures have attained promising accuracy on point cloud learning benchmarks compared to convolutional neural networks. However, existing point cloud Transformers are computationally expensive because they waste a significant amount of time on structuring irregular data. To solve this shortcoming, we present the Sparse Window Attention module to gather coarse‐grained local features from nonempty voxels. The module not only bypasses the expensive irregular data structuring and invalid empty voxel computation, but also obtains linear computational complexity with respect to voxel resolution. Meanwhile, we leverage two different self‐attention variants to gather fine‐grained features about the global shape according to different scale of point clouds. Finally, we construct our neural architecture called point‐voxel transformer (PVT), which integrates these modules into a joint framework for point cloud learning. Compared with previous transformer‐based and attention‐based models, our method attains a top accuracy of 94.1% on the classification benchmark and 10-0.25em× $10\times $ inference speedup on average. Extensive experiments also validate the effectiveness of PVT on semantic segmentation benchmarks. Our code and pretrained model are avaliable at https://github.com/HaochengWan/PVT.

show abstract

SnowflakeNet: Point Cloud Completion by Snowflake Point Deconvolution with Skip-Transformer

Cited by 157 publications

References 41 publications

Point cloud completion on structured feature map with feedback network

Point cloud completion on structured feature map with feedback network

SoftPool++: An Encoder–Decoder Network for Point Cloud Completion

PVT: Point‐voxel transformer for point cloud learning

Contact Info

Product

Resources

About