Abstract
Video titling and question answering are two important tasks toward high-level visual data understanding. To address those two tasks, we propose a large-scale dataset and demonstrate several models on such dataset in this work. A great video title describes the most salient event compactly and captures the viewer's attention. In contrast, video captioning tends to generate sentences that describe the video as a whole. Although generating a video title automatically is a very useful task, it is much less addressed than video captioning. We address video title generation for the first time by proposing two methods that extend state-of-the-art video captioners to this new task. First, we make video captioners highlight sensitive by priming them with a highlight detector. Our framework allows for jointly training a model for title generation and video highlight localization. Second, we induce high sentence diversity in video captioners, so that the generated titles are also diverse and catchy. This means that a large number of sentences might be required to learn the sentence structure of titles. Hence, we propose a novel sentence augmentation method to train a captioner with additional sentence-only examples that come without corresponding videos. On the other hand, for video question-answering task: we propose to learn a deep model to answer a free-form natural language question about the contents of a video. We make a program automatically harvests a large number of videos and descriptions freely available online. Then, a large number of candidate QA pairs are automatically generated from descriptions rather than manually annotated. Next, we use these candidate QA pairs to train a number of video-based QA methods extended from MN, VQA, SA, and SS. In order to handle non-perfect candidate QA pairs, we propose a self-paced learning procedure to iteratively identify them and mitigate their effects in training. To demonstrate our idea, we collected a large-scale Video Titles in the Wild (VTW) dataset of $18100$ automatically crawled user-generated videos and titles. We then utilize an automatic QA generator to generate a large number of QA pairs for training and collect the manually generated QA pairs from Amazon Mechanical Turk. On VTW, our methods consistently improve title prediction accuracy, and achieve the best performance in both automatic and human evaluation. Next, our sentence augmentation method also outperforms the baselines on the M-VAD dataset. Finally, the results of video question answering show that our self-paced learning procedure is effective, and the extended SS model outperforms various baselines.