Abstract
Many websites and applications that enable users to view videos include a feature that lets a user watching a video hover on the timestamp bar to see preview images of the content at or near that timestamp. However, selection of preview images that is unaware of the video content can cause the thumbnails to be non-representative and unappealing to the user. This disclosure describes the use of crowdsourced information on video pause actions performed by various users to select preview images for different parts of a video. An engagement prior model is built on the crowdsourced information. A temporal histogram of the video based on prior user pause actions enables intelligent preview image selection, e.g., using a gradient ascent algorithm. The thumbnails thus selected can provide a more satisfactory preview experience for the user. The temporal heatmap can be updated and preview image selection repeated as more viewers view the video and perform pause actions.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Shin, D, "Video Preview Image Selection Based on Prior User Interactions", Technical Disclosure Commons, (January 17, 2023)
https://www.tdcommons.org/dpubs_series/5641