Learning Aligned Cross-Modal Representations from Weakly Aligned Data

TL;DR

This paper introduces methods to learn shared, modality-agnostic scene representations across different data types, enabling better cross-modal transfer and retrieval, supported by a new dataset and visualization insights.

Contribution

It proposes regularization techniques for cross-modal CNNs to develop aligned, shared scene representations, advancing cross-modal transfer capabilities.

Findings

Shared representations improve cross-modal retrieval.

Units activate on consistent concepts regardless of modality.

Proposed methods outperform baseline models.

Abstract

People can recognize scenes across many different modalities beyond natural images. In this paper, we investigate how to learn cross-modal scene representations that transfer across modalities. To study this problem, we introduce a new cross-modal scene dataset. While convolutional neural networks can categorize cross-modal scenes well, they also learn an intermediate representation not aligned across modalities, which is undesirable for cross-modal transfer applications. We present methods to regularize cross-modal convolutional neural networks so that they have a shared representation that is agnostic of the modality. Our experiments suggest that our scene representation can help transfer representations across modalities for retrieval. Moreover, our visualizations suggest that units emerge in the shared representation that tend to activate on consistent concepts independently of the modality.