Corpus ID: 52195297

Parallel Separable 3D Convolution for Video and Volumetric Data Understanding

@article{Gonda2018ParallelS3,
  title={Parallel Separable 3D Convolution for Video and Volumetric Data Understanding},
  author={Felix Gonda and D. Wei and T. Parag and H. Pfister},
  journal={ArXiv},
  year={2018},
  volume={abs/1809.04096}
}
For video and volumetric data understanding, 3D convolution layers are widely used in deep learning, however, at the cost of increasing computation and training time. Recent works seek to replace the 3D convolution layer with convolution blocks, e.g. structured combinations of 2D and 1D convolution layers. In this paper, we propose a novel convolution block, Parallel Separable 3D Convolution (PmSCn), which applies m parallel streams of n 2D and one 1D convolution layers along different… Expand

Figures and Topics from this paper

Ghost-Light-3dnet: Efficient Network For Heart Segmentation
A multi-branch separable convolution neural network for pedestrian attribute recognition
HIVE-Net: Centerline-Aware HIerarchical View-Ensemble Convolutional Network for Mitochondria Segmentation in EM Images
VideoSSL: Semi-Supervised Learning for Video Classification
Reinventing 2D Convolutions for 3D Images.
CAKES: Channel-wise Automatic KErnel Shrinking for Efficient 3D Network

References

SHOWING 1-10 OF 49 REFERENCES
Learning Spatio-Temporal Representation with Pseudo-3D Residual Networks
Learning Spatiotemporal Features with 3D Convolutional Networks
Automatic 3D Cardiovascular MR Segmentation with Densely-Connected Volumetric ConvNets
Convolutional Two-Stream Network Fusion for Video Action Recognition
3D Densely Convolutional Networks for Volumetric Segmentation
Two-Stream Convolutional Networks for Action Recognition in Videos
Densely Connected Convolutional Networks
Spatiotemporal Residual Networks for Video Action Recognition
...
1
2
3
4
5
...