Partially-Supervised Novel Object Captioning Leveraging Context from Paired Data
This addresses the challenge of captioning novel objects in images for computer vision applications, representing an incremental advance by leveraging partially labeled data to enhance existing models.
The paper tackles the problem of generating captions for images containing novel objects without caption labels by proposing PS-NOC, a training approach that uses synthetic data from existing pairs and pseudo-labels, achieving an 85.9 F1-score and 103.8 CIDEr on an out-of-domain test split, which are improvements of 85.9 and 34.1 points over the baseline.
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.