Context-aware Captions from Context-agnostic Supervision

TL;DR

This paper presents a novel inference method to generate context-aware image captions that distinguish similar concepts using only generic, context-agnostic training data, improving discriminative captioning performance.

Contribution

It introduces a joint inference technique combining a context-agnostic language model with a listener for discriminative captioning, applicable without specialized training data.

Findings

Outperforms baseline methods in discriminative captioning tasks

Effective in distinguishing closely related visual concepts

Improves justification accuracy for fine-grained categories

Abstract

We introduce an inference technique to produce discriminative context-aware image captions (captions that describe differences between images or visual concepts) using only generic context-agnostic training data (captions that describe a concept or an image in isolation). For example, given images and captions of "siamese cat" and "tiger cat", we generate language that describes the "siamese cat" in a way that distinguishes it from "tiger cat". Our key novelty is that we show how to do joint inference over a language model that is context-agnostic and a listener which distinguishes closely-related concepts. We first apply our technique to a justification task, namely to describe why an image contains a particular fine-grained category as opposed to another closely-related category of the CUB-200-2011 dataset. We then study discriminative image captioning to generate language that uniquely refers to one of two semantically-similar images in the COCO dataset. Evaluations with discriminative ground truth for justification and human studies for discriminative image captioning reveal that our approach outperforms baseline generative and speaker-listener approaches for discrimination.