SketRet: Bi-Level Domain Adaptation for Zero-Shot Sketch-Based Image Retrieval

The efficacy of zero-shot sketch-based image retrieval (ZS-SBIR) models is governed by two challenges. The immense distributions-gap between the sketches and the images requires a proper domain alignment. Moreover, the fine-grained nature of the task and the high intra-class variance of many categories necessitates a class-wise discriminative mapping among the sketch, image, and the semantic spaces. Under this premise, we propose BDA-SketRet, a novel ZS-SBIR framework performing a bi-level domain adaptation for aligning the spatial and semantic features of the visual data pairs progressively. we propose a novel symmetric loss function based on the notion of information bottleneck for aligning the semantic features while a cross-entropy-based adversarial loss is introduced to align the spatial feature maps.

Please find the paper

If you find this work useful, please cite our paper

@article{CHAUDHURI2022245,
title = {BDA-SketRet: Bi-level domain adaptation for zero-shot SBIR},
journal = {Neurocomputing},
year = {2022},
author = {Ushasi Chaudhuri and Ruchika Chavan and Biplab Banerjee and Anjan Dutta and Zeynep Akata},
}