Sketch3T: Test-Time Training for Zero-Shot SBIR

TL;DR

This paper introduces Sketch3T, a test-time training approach for zero-shot sketch-based image retrieval that adapts to new categories and sketch styles using a self-supervised auxiliary task and meta-learning, outperforming existing methods.

Contribution

The paper proposes a novel test-time training paradigm with a self-supervised auxiliary task and meta-learning to adapt to new categories and sketch styles in zero-shot SBIR.

Findings

Outperforms state-of-the-art methods in zero-shot SBIR.

Effectively adapts to new sketching styles at test time.

Demonstrates robustness to distribution shifts in sketches.

Abstract

Zero-shot sketch-based image retrieval typically asks for a trained model to be applied as is to unseen categories. In this paper, we question to argue that this setup by definition is not compatible with the inherent abstract and subjective nature of sketches, i.e., the model might transfer well to new categories, but will not understand sketches existing in different test-time distribution as a result. We thus extend ZS-SBIR asking it to transfer to both categories and sketch distributions. Our key contribution is a test-time training paradigm that can adapt using just one sketch. Since there is no paired photo, we make use of a sketch raster-vector reconstruction module as a self-supervised auxiliary task. To maintain the fidelity of the trained cross-modal joint embedding during test-time update, we design a novel meta-learning based training paradigm to learn a separation between model updates incurred by this auxiliary task from those off the primary objective of discriminative learning. Extensive experiments show our model to outperform state of-the-arts, thanks to the proposed test-time adaption that not only transfers to new categories but also accommodates to new sketching styles.