End-to-end Generative Zero-shot Learning via Few-shot Learning

TL;DR

Z2FSL is an end-to-end framework that combines generative zero-shot learning with few-shot learning, jointly training both modules to improve ZSL performance across benchmarks.

Contribution

It introduces a novel joint training framework that integrates generative ZSL with FSL, effectively reducing ZSL to FSL and achieving state-of-the-art results.

Findings

Consistent improvement over baselines

State-of-the-art performance on benchmarks

Flexible integration of various FSL algorithms

Abstract

Contemporary state-of-the-art approaches to Zero-Shot Learning (ZSL) train generative nets to synthesize examples conditioned on the provided metadata. Thereafter, classifiers are trained on these synthetic data in a supervised manner. In this work, we introduce Z2FSL, an end-to-end generative ZSL framework that uses such an approach as a backbone and feeds its synthesized output to a Few-Shot Learning (FSL) algorithm. The two modules are trained jointly. Z2FSL solves the ZSL problem with a FSL algorithm, reducing, in effect, ZSL to FSL. A wide class of algorithms can be integrated within our framework. Our experimental results show consistent improvement over several baselines. The proposed method, evaluated across standard benchmarks, shows state-of-the-art or competitive performance in ZSL and Generalized ZSL tasks.