• Computer Science
  • Published in ArXiv 2018

ELASTIC: Improving CNNs with Instance Specific Scaling Policies

@article{Wang2018ELASTICIC,
  title={ELASTIC: Improving CNNs with Instance Specific Scaling Policies},
  author={Huiyu Wang and Aniruddha Kembhavi and Ali Farhadi and Alan L. Yuille and Mohammad Rastegari},
  journal={ArXiv},
  year={2018},
  volume={abs/1812.05262}
}
Scale variation has been a challenge from traditional to modern approaches in computer vision. Most solutions to scale issues have similar theme: a set of intuitive and manually designed policies that are generic and fixed (e.g. SIFT or feature pyramid). We argue that the scale policy should be learned from data. In this paper, we introduce ELASTIC, a simple, efficient and yet very effective approach to learn instance-specific scale policy from data. We formulate the scaling policy as a non… CONTINUE READING
5
Twitter Mentions

References

Publications referenced by this paper.
SHOWING 1-10 OF 40 REFERENCES

Microsoft COCO: Common Objects in Context

VIEW 5 EXCERPTS
HIGHLY INFLUENTIAL

Aggregated Residual Transformations for Deep Neural Networks

VIEW 4 EXCERPTS
HIGHLY INFLUENTIAL

Densely Connected Convolutional Networks

VIEW 5 EXCERPTS
HIGHLY INFLUENTIAL

CondenseNet: An Efficient DenseNet Using Learned Group Convolutions

VIEW 1 EXCERPT

Deep Layer Aggregation

VIEW 2 EXCERPTS

Dilated Residual Networks

VIEW 1 EXCERPT

Improving Pairwise Ranking for Multi-label Image Classification

VIEW 1 EXCERPT

Learning Spatial Regularization with Image-Level Supervisions for Multi-label Image Classification

VIEW 1 EXCERPT