Contextually Customized Video Summaries via Natural Language
This addresses the need for customizable video summaries for users with varying preferences, though it is incremental as it builds on existing embedding and summarization techniques.
The paper tackles the problem of generating personalized video summaries based on natural language descriptions, achieving performance comparable to or exceeding baseline methods that use ground-truth information.
The best summary of a long video differs among different people due to its highly subjective nature. Even for the same person, the best summary may change with time or mood. In this paper, we introduce the task of generating customized video summaries through simple text. First, we train a deep architecture to effectively learn semantic embeddings of video frames by leveraging the abundance of image-caption data via a progressive and residual manner. Given a user-specific text description, our algorithm is able to select semantically relevant video segments and produce a temporally aligned video summary. In order to evaluate our textually customized video summaries, we conduct experimental comparison with baseline methods that utilize ground-truth information. Despite the challenging baselines, our method still manages to show comparable or even exceeding performance. We also show that our method is able to generate semantically diverse video summaries by only utilizing the learned visual embeddings.