Resolution learning in deep convolutional networks using scale-space theory

Abstract

Resolution in deep convolutional neural networks (CNNs) is typically bounded by the receptive field size through filter sizes, and subsampling layers or strided convolutions on feature maps. The optimal resolution may vary significantly depending on the dataset. Modern CNNs hard-code their resolution hyper-parameters in the network architecture which makes tuning such hyper-parameters cumbersome. We propose to do away with hard-coded resolution hyper-parameters and aim to learn the appropriate resolution from data. We use scale-space theory to obtain a self-similar parametrization of filters and make use of the N-Jet: a truncated Taylor series to approximate a filter by a learned combination of Gaussian derivative filters. The parameter sigma of the Gaussian basis controls both the amount of detail the filter encodes and the spatial extent of the filter. Since sigma is a continuous parameter, we can optimize it with respect to the loss. The proposed N-Jet layer achieves comparable performance when used in state-of-the art architectures, while learning the correct resolution in each layer automatically. We evaluate our N-Jet layer on both classification and segmentation, and we show that learning sigma is especially beneficial for inputs at multiple sizes.

Preprint accepted by IEEE Transactions on Image Processing, 2021 (TIP). Link to final published article: https://ieeexplore.ieee.org/abstract/document/9552550

References in corpus (10)

• Rethinking Atrous Convolution for Semantic Image Segmentation
• Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms
• Understanding the Effective Receptive Field in Deep Convolutional Neural Networks
• Stochastic Pooling for Regularization of Deep Convolutional Neural Networks
• Making Convolutional Networks Shift-Invariant Again
• Fractional Max-Pooling
• Spectral Representations for Convolutional Neural Networks
• Locally Scale-Invariant Convolutional Neural Networks
• Blurring the Line Between Structure and Learning to Optimize and Adapt Receptive Fields
• From Discrete to Continuous Convolution Layers

Cited by in corpus (5)

• LDConv: Linear deformable convolution for improving convolutional neural networks
• Discrete approximations of Gaussian smoothing and Gaussian derivatives
• Covariance properties under natural image transformations for the generalized Gaussian derivative model for visual receptive fields
• Scale generalisation properties of extended scale-covariant and scale-invariant Gaussian derivative networks on image datasets with spatial scaling variations
• Approximation properties relative to continuous scale space for hybrid discretizations of Gaussian derivative operators