Depthwise Convolution is All You Need for Learning Multiple Visual Domains

TL;DR

This paper introduces a multi-domain learning model using depthwise separable convolution that efficiently captures shared structures across visual domains, achieving high performance with fewer parameters.

Contribution

The paper proposes a novel multi-domain learning architecture based on depthwise separable convolution with a gating mechanism for soft sharing, reducing parameters and improving performance.

Findings

Achieves highest score on Visual Decathlon Challenge

Uses only 50% of parameters compared to state-of-the-art

Effectively captures shared cross-domain features

Abstract

There is a growing interest in designing models that can deal with images from different visual domains. If there exists a universal structure in different visual domains that can be captured via a common parameterization, then we can use a single model for all domains rather than one model per domain. A model aware of the relationships between different domains can also be trained to work on new domains with less resources. However, to identify the reusable structure in a model is not easy. In this paper, we propose a multi-domain learning architecture based on depthwise separable convolution. The proposed approach is based on the assumption that images from different domains share cross-channel correlations but have domain-specific spatial correlations. The proposed model is compact and has minimal overhead when being applied to new domains. Additionally, we introduce a gating mechanism to promote soft sharing between different domains. We evaluate our approach on Visual Decathlon Challenge, a benchmark for testing the ability of multi-domain models. The experiments show that our approach can achieve the highest score while only requiring 50% of the parameters compared with the state-of-the-art approaches.