Shashank Bujimalla

Shashank Bujimalla, Mahesh Subedar, Omesh Tickoo

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.

Shashank Bujimalla, Mahesh Subedar, Omesh Tickoo

In this paper, we study the impact of motion blur, a common quality flaw in real world images, on a state-of-the-art two-stage image captioning solution, and notice a degradation in solution performance as blur intensity increases. We investigate techniques to improve the robustness of the solution to motion blur using training data augmentation at each or both stages of the solution, i.e., object detection and captioning, and observe improved results. In particular, augmenting both the stages reduces the CIDEr-D degradation for high motion blur intensity from 68.7 to 11.7 on MS COCO dataset, and from 22.4 to 6.8 on Vizwiz dataset.

Shashank Bujimalla, Mahesh Subedar, Omesh Tickoo

Bayesian deep neural networks (DNNs) can provide a mathematically grounded framework to quantify uncertainty in predictions from image captioning models. We propose a Bayesian variant of policy-gradient based reinforcement learning training technique for image captioning models to directly optimize non-differentiable image captioning quality metrics such as CIDEr-D. We extend the well-known Self-Critical Sequence Training (SCST) approach for image captioning models by incorporating Bayesian inference, and refer to it as B-SCST. The "baseline" for the policy-gradients in B-SCST is generated by averaging predictive quality metrics (CIDEr-D) of the captions drawn from the distribution obtained using a Bayesian DNN model. We infer this predictive distribution using Monte Carlo (MC) dropout approximate variational inference. We show that B-SCST improves CIDEr-D scores on Flickr30k, MS COCO and VizWiz image captioning datasets, compared to the SCST approach. We also provide a study of uncertainty quantification for the predicted captions, and demonstrate that it correlates well with the CIDEr-D scores. To our knowledge, this is the first such analysis, and it can improve the interpretability of image captioning model outputs, which is critical for practical applications.