Abstract
Without any doubt, state of art advertisement insertion mechanisms along with the requested video content has emerged to be the most crucial part of online video delivery ecosystems. There are several widely deployed methods which have been used since the first days of video streaming such as tag word matching between video content versus advertisement content along with manual matching-based approaches. Conventional but non-scalable and context independent methods cannot fulfil the requirements of an online video platform when there are several millions of user generated videos along with premium content and advertisement of varying production quality. In such environment, a content aware advertisement insertion framework is required based on object recognition, machine learning and artificial intelligence to understand the context of the video and match appropriate advertisement and stitch the advertisement at the most convenient moment of the target video content. In this paper, SuperEye; a deployment ready, content aware, scalable, distributed advertisement insertion framework for a 5G oriented online video platform is designed and developed. The foundational object analyzing mechanism of the underlying system examine each particular context that is part of the wider video and advertisement catalogue using object recognition while generating a time-lapse map of all objects that are detected through the video. Based upon this information, the framework matches the most significant object that is detected for a particular interval and associates the advertisement with similar properties. Additionally, this novel technique does not require any watch history or personalized data related to the user, but primarily interested in only the current requested content information. Therefore, this framework can work along with any type of recommendation engine or rank based association algorithm. The proposed framework is independent of the user information and regarding the subjective user results collected, successful video to ad match ratios of SuperEye significantly exceed the current implementations of YouTube, Vimeo and DailyMotion.
| Original language | English |
|---|---|
| Pages (from-to) | 9361-9379 |
| Number of pages | 19 |
| Journal | Multimedia Tools and Applications |
| Volume | 82 |
| Issue number | 6 |
| DOIs | |
| Publication status | Published - 3 Sept 2022 |
Bibliographical note
Publisher Copyright:© 2022, The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.
Keywords
- Advertisement insertion
- Artificial intelligence
- Behavior detection
- Object detection
- Object recognition
- Online video streaming
- Quality of experience