Improving the Resolution of CNN Feature Maps Efficiently with Multisampling

Abstract

We describe a new class of subsampling techniques for CNNs, termed multisampling, that significantly increases the amount of information kept by feature maps through subsampling layers. One version of our method, which we call checkered subsampling, significantly improves the accuracy of state-of-the-art architectures such as DenseNet and ResNet without any additional parameters and, remarkably, improves the accuracy of certain pretrained ImageNet models without any training or fine-tuning. We glean possible insight into the nature of data augmentations and demonstrate experimentally that coarse feature maps are bottlenecking the performance of neural networks in image classification.

References in corpus (12)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Rethinking Atrous Convolution for Semantic Image Segmentation
• Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation
• Densely Connected Convolutional Networks
• Multi-Scale Context Aggregation by Dilated Convolutions
• DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs
• Stochastic Pooling for Regularization of Deep Convolutional Neural Networks
• Training Deep Nets with Sublinear Memory Cost
• Deformable Convolutional Networks
• Fractional Max-Pooling
• Memory-Efficient Implementation of DenseNets
• Deep Pyramidal Residual Networks