Partially-Supervised Novel Object Captioning Leveraging Context from Paired Data

TL;DR

This paper introduces PS-NOC, a training approach for image captioning that effectively incorporates novel objects without caption labels, achieving state-of-the-art results on out-of-domain MS COCO data.

Contribution

The paper presents a novel training method for captioning models that leverages partially paired data and synthetic captions for novel objects, improving out-of-domain performance.

Findings

Achieved 85.9 F1-score on novel objects, surpassing baseline by 85.9 points.

Improved CIDEr score to 103.8, a 34.1 point increase over baseline.

Demonstrated effectiveness through extensive ablation studies.

Abstract

In this paper, we propose an approach to improve image captioning solution for images with novel objects that do not have caption labels in the training dataset. We refer to our approach as Partially-Supervised Novel Object Captioning (PS-NOC). PS-NOC is agnostic to model architecture, and primarily focuses on the training approach that uses existing fully paired image-caption data and the images with only the novel object detection labels (partially paired data). We create synthetic paired captioning data for novel objects by leveraging context from existing image-caption pairs. We then create pseudo-label captions for partially paired images with novel objects, and use this additional data to fine-tune the captioning model. We also propose a variant of SCST within PS-NOC, called SCST-F1, that directly optimizes the F1-score of novel objects. Using a popular captioning model (Up-Down) as baseline, PS-NOC sets new state-of-the-art results on held-out MS COCO out-of-domain test split, i.e., 85.9 F1-score and 103.8 CIDEr. This is an improvement of 85.9 and 34.1 points respectively compared to baseline model that does not use partially paired data during training. We also perform detailed ablation studies to demonstrate the effectiveness of our approach.