Few-Shot Inspired Generative Zero-Shot Learning

TL;DR

This paper introduces FSIGenZ, a novel few-shot inspired generative zero-shot learning framework that reduces synthetic data reliance by dynamically re-scoring attributes and using class prototypes, achieving competitive results.

Contribution

The paper proposes MSAS for dynamic attribute re-scoring and a prototype-based approach, reducing synthetic data needs in ZSL.

Findings

Achieves competitive performance with fewer synthetic features.

Demonstrates effectiveness on SUN, AwA2, and CUB benchmarks.

Introduces a semantic-aware contrastive classifier.

Abstract

Generative zero-shot learning (ZSL) methods typically synthesize visual features for unseen classes using predefined semantic attributes, followed by training a fully supervised classification model. While effective, these methods require substantial computational resources and extensive synthetic data, thereby relaxing the original ZSL assumptions. In this paper, we propose FSIGenZ, a few-shot-inspired generative ZSL framework that reduces reliance on large-scale feature synthesis. Our key insight is that class-level attributes exhibit instance-level variability, i.e., some attributes may be absent or partially visible, yet conventional ZSL methods treat them as uniformly present. To address this, we introduce Model-Specific Attribute Scoring (MSAS), which dynamically re-scores class attributes based on model-specific optimization to approximate instance-level variability without access to unseen data. We further estimate group-level prototypes as clusters of instances based on MSAS-adjusted attribute scores, which serve as representative synthetic features for each unseen class. To mitigate the resulting data imbalance, we introduce a Dual-Purpose Semantic Regularization (DPSR) strategy while training a semantic-aware contrastive classifier (SCC) using these prototypes. Experiments on SUN, AwA2, and CUB benchmarks demonstrate that FSIGenZ achieves competitive performance using far fewer synthetic features.