Cross-Modality Distillation: A case for Conditional Generative Adversarial Networks

TL;DR

This paper introduces a novel use of Conditional Generative Adversarial Networks to transfer knowledge across sensor modalities, improving low-resolution target detection in noisy, incomplete surveillance data.

Contribution

It proposes a CGAN-based method for cross-modality knowledge distillation, effectively handling missing modalities and outperforming traditional and recent teacher-student models.

Findings

CGAN-based distillation improves detection accuracy.

Method handles missing and noisy sensor data effectively.

Outperforms traditional approaches and recent models.

Abstract

In this paper, we propose to use a Conditional Generative Adversarial Network (CGAN) for distilling (i.e. transferring) knowledge from sensor data and enhancing low-resolution target detection. In unconstrained surveillance settings, sensor measurements are often noisy, degraded, corrupted, and even missing/absent, thereby presenting a significant problem for multi-modal fusion. We therefore specifically tackle the problem of a missing modality in our attempt to propose an algorithm based on CGANs to generate representative information from the missing modalities when given some other available modalities. Despite modality gaps, we show that one can distill knowledge from one set of modalities to another. Moreover, we demonstrate that it achieves better performance than traditional approaches and recent teacher-student models.