对于一家在线视频服务公司来讲,理解视频的内容其重要性不言而喻,只有深度理解用户观看的内容到底是什么,才能更好的给用户提供个性化的内容推荐、更好的交互体验等产品服务。
Hulu自2016年开始系统性的在视频内容理解方面展开研究,从视频切分、人工合成元素抽取、视频标签生成、精彩片段分析等等课题入手,通过构建系统平台来支撑视频数据的生成和处理,并对业务及产品的支持方面也多有探索。这其中也积累了一些经验,期望借助这个平台的分享,和大家交流Hulu在这个领域是如何探索和应用的,共同探索这个领域的未来发展趋势。
演讲提纲:
We will cover :
- Importance and urgency of doing video content understanding in Hulu
- Three main research directions of video understanding in Hulu
- Automation Tools
- Video derived tags, tag lake and tag governance
- Content generation
- AI platform's architecture and pipeline for video understanding
- FrameHouse
- ML/DL platform
- Automation pipeline and architecture
- Data management and serving
- Business and product support and best practice
- Content embedding and deep personalization
- Ads related experience
- UX innovation
听众受益点:
- Understand the whole pipeline and architecture of video understanding in Hulu, and learn how Hulu enable AI algorithms through AI platform
- Learn the best practice of how Hulu leveraging video understanding techs to support business and product innovation
- How Hulu process video derived tags and do tag governance