Adversarial Knowledge Transfer from Unlabeled Data

TL;DR

This paper introduces an adversarial knowledge transfer framework that leverages unlabeled internet data to enhance visual recognition, especially when labeled data is scarce or unavailable.

Contribution

It proposes a novel adversarial learning method that aligns feature spaces between unlabeled source data and labeled target data without requiring class overlap or pretext tasks.

Findings

Improves classifier performance using unlabeled data.

Effective across multiple visual recognition tasks.

No need for class overlap or pretext tasks.

Abstract

While machine learning approaches to visual recognition offer great promise, most of the existing methods rely heavily on the availability of large quantities of labeled training data. However, in the vast majority of real-world settings, manually collecting such large labeled datasets is infeasible due to the cost of labeling data or the paucity of data in a given domain. In this paper, we present a novel Adversarial Knowledge Transfer (AKT) framework for transferring knowledge from internet-scale unlabeled data to improve the performance of a classifier on a given visual recognition task. The proposed adversarial learning framework aligns the feature space of the unlabeled source data with the labeled target data such that the target classifier can be used to predict pseudo labels on the source data. An important novel aspect of our method is that the unlabeled source data can be of different classes from those of the labeled target data, and there is no need to define a separate pretext task, unlike some existing approaches. Extensive experiments well demonstrate that models learned using our approach hold a lot of promise across a variety of visual recognition tasks on multiple standard datasets.