Semi Supervised Phrase Localization in a Bidirectional Caption-Image Retrieval Framework

TL;DR

This paper presents a deep neural network that learns to localize phrases in images using semi-supervised training, leveraging a joint embedding space for improved retrieval and localization without explicit supervision.

Contribution

The authors introduce a novel architecture that inherently learns localization from a bidirectional retrieval objective using a joint embedding space, outperforming existing methods.

Findings

Outperforms state-of-the-art in semi-supervised phrase localization

Effectively learns localization without explicit supervision

Achieves strong retrieval performance on MSCOCO and Flickr30K datasets

Abstract

We introduce a novel deep neural network architecture that links visual regions to corresponding textual segments including phrases and words. To accomplish this task, our architecture makes use of the rich semantic information available in a joint embedding space of multi-modal data. From this joint embedding space, we extract the associative localization maps that develop naturally, without explicitly providing supervision during training for the localization task. The joint space is learned using a bidirectional ranking objective that is optimized using a $N$-Pair loss formulation. This training mechanism demonstrates the idea that localization information is learned inherently while optimizing a Bidirectional Retrieval objective. The model's retrieval and localization performance is evaluated on MSCOCO and Flickr30K Entities datasets. This architecture outperforms the state of the art results in the semi-supervised phrase localization setting.