Parallel Grid Pooling for Data Augmentation

TL;DR

The paper introduces parallel grid pooling (PGP), a novel layer for CNNs that performs downsampling without losing features, acting as a data augmentation method and improving image classification performance.

Contribution

It proposes PGP, a new downsampling layer that preserves all features and enhances CNNs as a data augmentation technique, also linking dilated convolution to PGP.

Findings

PGP improves classification accuracy on benchmark datasets.

Dilated convolution can be represented using PGP operations.

PGP complements existing data augmentation methods.

Abstract

Convolutional neural network (CNN) architectures utilize downsampling layers, which restrict the subsequent layers to learn spatially invariant features while reducing computational costs. However, such a downsampling operation makes it impossible to use the full spectrum of input features. Motivated by this observation, we propose a novel layer called parallel grid pooling (PGP) which is applicable to various CNN models. PGP performs downsampling without discarding any intermediate feature. It works as data augmentation and is complementary to commonly used data augmentation techniques. Furthermore, we demonstrate that a dilated convolution can naturally be represented using PGP operations, which suggests that the dilated convolution can also be regarded as a type of data augmentation technique. Experimental results based on popular image classification benchmarks demonstrate the effectiveness of the proposed method. Code is available at: https://github.com/akitotakeki