A causal view of compositional zero-shot recognition

TL;DR

This paper introduces a causal approach to compositional zero-shot recognition, enabling models to better generalize to new attribute-object combinations by learning disentangled, causally-inspired representations.

Contribution

It proposes a causal perspective for zero-shot learning and develops a causal-inspired embedding model that improves compositional generalization.

Findings

Improved accuracy on synthesized attribute-object datasets

Enhanced generalization to unseen attribute-object pairs

Causal-inspired embeddings outperform baselines

Abstract

People easily recognize new visual categories that are new combinations of known components. This compositional generalization capacity is critical for learning in real-world domains like vision and language because the long tail of new combinations dominates the distribution. Unfortunately, learning systems struggle with compositional generalization because they often build on features that are correlated with class labels even if they are not "essential" for the class. This leads to consistent misclassification of samples from a new distribution, like new combinations of known components. Here we describe an approach for compositional generalization that builds on causal ideas. First, we describe compositional zero-shot learning from a causal perspective, and propose to view zero-shot inference as finding "which intervention caused the image?". Second, we present a causal-inspired embedding model that learns disentangled representations of elementary components of visual objects from correlated (confounded) training data. We evaluate this approach on two datasets for predicting new combinations of attribute-object pairs: A well-controlled synthesized images dataset and a real-world dataset which consists of fine-grained types of shoes. We show improvements compared to strong baselines.