Using Descriptive Video Services to Create a Large Data Source for Video Annotation Research
This provides a valuable resource for video description research, though it is incremental as it builds on existing DVS content with a new collection method.
The authors tackled the problem of creating a large-scale dataset for video annotation by automatically extracting and aligning Descriptive Video Service (DVS) narrations from DVDs, resulting in a dataset of over 84.6 hours of paired video and natural language descriptions from 92 DVDs.
In this work, we introduce a dataset of video annotated with high quality natural language phrases describing the visual content in a given segment of time. Our dataset is based on the Descriptive Video Service (DVS) that is now encoded on many digital media products such as DVDs. DVS is an audio narration describing the visual elements and actions in a movie for the visually impaired. It is temporally aligned with the movie and mixed with the original movie soundtrack. We describe an automatic DVS segmentation and alignment method for movies, that enables us to scale up the collection of a DVS-derived dataset with minimal human intervention. Using this method, we have collected the largest DVS-derived dataset for video description of which we are aware. Our dataset currently includes over 84.6 hours of paired video/sentences from 92 DVDs and is growing.