OneDConv: Generalized Convolution For Transform-Invariant Representation

TL;DR

This paper introduces OneDConv, a generalized convolution operator that enhances transform-invariance in CNNs, improving robustness and generalization without performance loss, and can be integrated into existing architectures.

Contribution

The paper proposes a novel dynamic convolution operator, OneDConv, that achieves transform-invariance efficiently and can replace standard convolutions in CNNs.

Findings

Outperforms vanilla convolution on benchmarks

Enhances robustness to image distortions

Maintains performance on standard images

Abstract

Convolutional Neural Networks (CNNs) have exhibited their great power in a variety of vision tasks. However, the lack of transform-invariant property limits their further applications in complicated real-world scenarios. In this work, we proposed a novel generalized one dimension convolutional operator (OneDConv), which dynamically transforms the convolution kernels based on the input features in a computationally and parametrically efficient manner. The proposed operator can extract the transform-invariant features naturally. It improves the robustness and generalization of convolution without sacrificing the performance on common images. The proposed OneDConv operator can substitute the vanilla convolution, thus it can be incorporated into current popular convolutional architectures and trained end-to-end readily. On several popular benchmarks, OneDConv outperforms the original convolution operation and other proposed models both in canonical and distorted images.