Skip to content

采用时间序列模型作为大语言模型的替代方案 #1

Description

@DanielZhangyc

考虑到使用LLM作为检测方案可能存在一定API开销以及使用门槛,且可能存在处理时间过长的问题。由于b站的视频都提供了视频不同时间段观看数据,而对于口播广告时间段可能存在一定明显的特征,可否采集视频的播放数据并使用时间序列模型(LSTM/Transformer)来进行广告片段的识别?

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions