Detail-Preserving Pooling in Deep Networks

TL;DR

This paper introduces detail-preserving pooling (DPP), an adaptive method inspired by the human visual system that enhances important structural details during downscaling in deep networks, leading to improved performance.

Contribution

The paper proposes a novel learnable pooling method, DPP, which magnifies spatial changes and preserves structural details, outperforming traditional pooling techniques.

Findings

DPP consistently outperforms max and average pooling on multiple datasets.

DPP's parameters can be learned jointly with the network.

Theoretical analysis supports DPP's ability to preserve important features.

Abstract

Most convolutional neural networks use some method for gradually downscaling the size of the hidden layers. This is commonly referred to as pooling, and is applied to reduce the number of parameters, improve invariance to certain distortions, and increase the receptive field size. Since pooling by nature is a lossy process, it is crucial that each such layer maintains the portion of the activations that is most important for the network's discriminability. Yet, simple maximization or averaging over blocks, max or average pooling, or plain downsampling in the form of strided convolutions are the standard. In this paper, we aim to leverage recent results on image downscaling for the purposes of deep learning. Inspired by the human visual system, which focuses on local spatial changes, we propose detail-preserving pooling (DPP), an adaptive pooling method that magnifies spatial changes and preserves important structural detail. Importantly, its parameters can be learned jointly with the rest of the network. We analyze some of its theoretical properties and show its empirical benefits on several datasets and networks, where DPP consistently outperforms previous pooling approaches.