Striving for Simplicity: The All Convolutional Net

TL;DR

This paper demonstrates that a simplified all convolutional network, replacing pooling with strided convolutions, can achieve competitive or state-of-the-art results in object recognition tasks, challenging traditional CNN design principles.

Contribution

The authors propose a new all convolutional architecture that omits pooling layers, replacing them with strided convolutions, and introduce a novel visualization method for CNN features.

Findings

Replacing max-pooling with strided convolutions maintains accuracy.

The all convolutional network achieves competitive or state-of-the-art results.

A new feature visualization technique applicable to various CNN structures.

Abstract

Most modern convolutional neural networks (CNNs) used for object recognition are built using the same principles: Alternating convolution and max-pooling layers followed by a small number of fully connected layers. We re-evaluate the state of the art for object recognition from small images with convolutional networks, questioning the necessity of different components in the pipeline. We find that max-pooling can simply be replaced by a convolutional layer with increased stride without loss in accuracy on several image recognition benchmarks. Following this finding -- and building on other recent work for finding simple network structures -- we propose a new architecture that consists solely of convolutional layers and yields competitive or state of the art performance on several object recognition datasets (CIFAR-10, CIFAR-100, ImageNet). To analyze the network we introduce a new variant of the "deconvolution approach" for visualizing features learned by CNNs, which can be applied to a broader range of network structures than existing approaches.