English

Shape Adaptor: A Learnable Resizing Module

Machine Learning 2020-08-11 v2 Computer Vision and Pattern Recognition Machine Learning

Abstract

We present a novel resizing module for neural networks: shape adaptor, a drop-in enhancement built on top of traditional resizing layers, such as pooling, bilinear sampling, and strided convolution. Whilst traditional resizing layers have fixed and deterministic reshaping factors, our module allows for a learnable reshaping factor. Our implementation enables shape adaptors to be trained end-to-end without any additional supervision, through which network architectures can be optimised for each individual task, in a fully automated way. We performed experiments across seven image classification datasets, and results show that by simply using a set of our shape adaptors instead of the original resizing layers, performance increases consistently over human-designed networks, across all datasets. Additionally, we show the effectiveness of shape adaptors on two other applications: network compression and transfer learning. The source code is available at: https://github.com/lorenmt/shape-adaptor.

Keywords

Cite

@article{arxiv.2008.00892,
  title  = {Shape Adaptor: A Learnable Resizing Module},
  author = {Shikun Liu and Zhe Lin and Yilin Wang and Jianming Zhang and Federico Perazzi and Edward Johns},
  journal= {arXiv preprint arXiv:2008.00892},
  year   = {2020}
}

Comments

Published at ECCV 2020