Spatially Adaptive Inference with Stochastic Feature Sampling and Interpolation

@article{Xie2020SpatiallyAI,
  title={Spatially Adaptive Inference with Stochastic Feature Sampling and Interpolation},
  author={Zhenda Xie and Zheng Zhang and X. Zhu and Gao Huang and Stephen Lin},
  journal={ArXiv},
  year={2020},
  volume={abs/2003.08866}
}
In the feature maps of CNNs, there commonly exists considerable spatial redundancy that leads to much repetitive processing. Towards reducing this superfluous computation, we propose to compute features only at sparsely sampled locations, which are probabilistically chosen according to activation responses, and then densely reconstruct the feature map with an efficient interpolation procedure. With this sampling-interpolation scheme, our network avoids expending computation on spatial locations… Expand
Glance and Focus: a Dynamic Approach to Reducing Spatial Redundancy in Image Classification
Adaptive Structured Sparse Network for Efficient CNNs with Feature Regularization
SegBlocks: Block-Based Dynamic Resolution Networks for Real-Time Segmentation
QueryDet: Cascaded Sparse Query for Accelerating High-Resolution Small Object Detection
  • Chenhongyi Yang, Zehao Huang, Naiyan Wang
  • Computer Science
  • ArXiv
  • 2021
CoDiNet: Path Distribution Modeling with Consistency and Diversity for Dynamic Routing.
Adaptive Focus for Efficient Video Recognition

References

SHOWING 1-10 OF 42 REFERENCES
SBNet: Sparse Blocks Network for Fast Inference
Spatially Adaptive Computation Time for Residual Networks
Stochastic Downsampling for Cost-Adjustable Inference and Improved Regularization in Convolutional Networks
Pruning Filters for Efficient ConvNets
Filter Pruning via Geometric Median for Deep Convolutional Neural Networks Acceleration
Rethinking the Smaller-Norm-Less-Informative Assumption in Channel Pruning of Convolution Layers
SeerNet: Predicting Convolutional Neural Network Feature-Map Sparsity Through Low-Bit Quantization
Importance Estimation for Neural Network Pruning
Centripetal SGD for Pruning Very Deep Convolutional Networks With Complicated Structure
Deformable Convolutional Networks
...
1
2
3
4
5
...