Dilated Convolution with Learnable Spacings: beyond bilinear interpolation

TL;DR

This paper enhances dilated convolution with learnable spacings by exploring advanced interpolation methods like Gaussian, improving image classification performance without increasing model complexity.

Contribution

It introduces longer-range interpolation techniques, including Gaussian interpolation, into DCLS, surpassing bilinear interpolation for better accuracy on ImageNet classification.

Findings

Gaussian interpolation improves accuracy over bilinear.

Longer-range interpolations enhance performance.

No increase in model parameters needed.

Abstract

Dilated Convolution with Learnable Spacings (DCLS) is a recently proposed variation of the dilated convolution in which the spacings between the non-zero elements in the kernel, or equivalently their positions, are learnable. Non-integer positions are handled via interpolation. Thanks to this trick, positions have well-defined gradients. The original DCLS used bilinear interpolation, and thus only considered the four nearest pixels. Yet here we show that longer range interpolations, and in particular a Gaussian interpolation, allow improving performance on ImageNet1k classification on two state-of-the-art convolutional architectures (ConvNeXt and Conv\-Former), without increasing the number of parameters. The method code is based on PyTorch and is available at https://github.com/K-H-Ismail/Dilated-Convolution-with-Learnable-Spacings-PyTorch