Deformable Convolution Module with Globally Learned Relative Offsets for Fundus Vessel Segmentation

TL;DR

This paper introduces a novel deformable convolution module with globally learned offsets, improving the capture of complex, long-distance features in fundus vessel segmentation, leading to state-of-the-art results.

Contribution

A new deformable convolutional module that learns subpixel displacements and captures global features, integrated into GDCUnet for superior fundus vessel segmentation.

Findings

GDCUnet achieves state-of-the-art performance on public datasets.

The deformable module enhances the learning of complex vessel features.

Ablation studies confirm the module's effectiveness in feature learning.

Abstract

Deformable convolution can adaptively change the shape of convolution kernel by learning offsets to deal with complex shape features. We propose a novel plug and play deformable convolutional module that uses attention and feedforward networks to learn offsets, so that the deformable patterns can capture long-distance global features. Compared with previously existing deformable convolutions, the proposed module learns the sub pixel displacement field and adaptively warps the feature maps across all channels rather than directly deforms the convolution kernel , which is equivalent to a relative deformation of the kernel sampling grids, achieving global feature deformation and the decoupling of kernel size and learning network. Considering that the fundus blood vessels have globally self similar complex edges, we design a deep learning model for fundus blood vessel segmentation, GDCUnet, based on the proposed convolutional module. Empirical evaluations under the same configuration and unified framework show that GDCUnet has achieved state of the art performance on public datasets. Further ablation experiments demonstrated that the proposed deformable convolutional module could more significantly learn the complex features of fundus blood vessels, enhancing the model representation and generalization capabilities. The proposed module is similar to the interface of conventional convolution, we suggest applying it to more machine vision tasks with complex global self similar features.