Dual Complementary Dynamic Convolution for Image Recognition

TL;DR

This paper introduces a novel dual complementary dynamic convolution (DCDC) operator that models scene features as local adaptive and global shared parts, significantly enhancing CNNs' representation capacity for image recognition tasks.

Contribution

The paper proposes the DCDC operator, a new dynamic convolution method that captures both local and global features, overcoming limitations of existing convolutions.

Findings

DCDC-ResNets outperform vanilla ResNets and other dynamic CNNs in image classification.

DCDC improves performance on object detection and segmentation tasks.

DCDC achieves higher accuracy with fewer FLOPs and parameters.

Abstract

As a powerful engine, vanilla convolution has promoted huge breakthroughs in various computer tasks. However, it often suffers from sample and content agnostic problems, which limits the representation capacities of the convolutional neural networks (CNNs). In this paper, we for the first time model the scene features as a combination of the local spatial-adaptive parts owned by the individual and the global shift-invariant parts shared to all individuals, and then propose a novel two-branch dual complementary dynamic convolution (DCDC) operator to flexibly deal with these two types of features. The DCDC operator overcomes the limitations of vanilla convolution and most existing dynamic convolutions who capture only spatial-adaptive features, and thus markedly boosts the representation capacities of CNNs. Experiments show that the DCDC operator based ResNets (DCDC-ResNets) significantly outperform vanilla ResNets and most state-of-the-art dynamic convolutional networks on image classification, as well as downstream tasks including object detection, instance and panoptic segmentation tasks, while with lower FLOPs and parameters.