Domain-Smoothing Network for Zero-Shot Sketch-Based Image Retrieval

TL;DR

This paper introduces a Domain-Smoothing Network that effectively reduces domain gap and intra-class diversity in zero-shot sketch-based image retrieval, significantly improving retrieval accuracy.

Contribution

The paper proposes a novel cross-modal contrastive learning approach and a category-specific memory bank to enhance ZS-SBIR performance, addressing key challenges neglected by prior methods.

Findings

Outperforms state-of-the-art on Sketchy and TU-Berlin datasets

Effectively reduces domain gap between sketches and images

Mitigates intra-class diversity in sketch representations

Abstract

Zero-Shot Sketch-Based Image Retrieval (ZS-SBIR) is a novel cross-modal retrieval task, where abstract sketches are used as queries to retrieve natural images under zero-shot scenario. Most existing methods regard ZS-SBIR as a traditional classification problem and employ a cross-entropy or triplet-based loss to achieve retrieval, which neglect the problems of the domain gap between sketches and natural images and the large intra-class diversity in sketches. Toward this end, we propose a novel Domain-Smoothing Network (DSN) for ZS-SBIR. Specifically, a cross-modal contrastive method is proposed to learn generalized representations to smooth the domain gap by mining relations with additional augmented samples. Furthermore, a category-specific memory bank with sketch features is explored to reduce intra-class diversity in the sketch domain. Extensive experiments demonstrate that our approach notably outperforms the state-of-the-art methods in both Sketchy and TU-Berlin datasets. Our source code is publicly available at https://github.com/haowang1992/DSN.