videoactivespeakerdetectmapper
July 17, 2026 · View on GitHub
Detect active speakers in a video by analyzing visual face tracks and audio signals, including consistency checks for gender and age. This operator uses the Light-ASD model to determine which tracked persons are actively speaking. It supports optional consistency detection using prior metadata from other operators. It must be operated after video_human_tracks_extraction_mapper and video_tagging_from_audio_mapper.
通过分析视觉人脸轨迹和音频信号来检测视频中的活跃说话者，包括性别和年龄的一致性检查。此算子使用 Light-ASD 模型来确定哪些被跟踪的人物正在主动说话。它支持使用其他算子的先验元数据进行可选的一致性检测。它必须在 video_human_tracks_extraction_mapper 和 video_tagging_from_audio_mapper 之后运行。
Type 算子类型: mapper
Tags 标签: gpu, video
🔧 Parameter Configuration 参数配置
| name 参数名 | type 类型 | default 默认值 | desc 说明 |
|---|---|---|---|
temp_save_path | <class 'str'> | './temp_path' | path for temporary file storage. |
Light_ASD_model_path | <class 'str'> | './thirdparty/humanvbench_models/Light-ASD/weight/finetuning_TalkSet.model' | path to the Light-ASD model weights. |
active_threshold | <class 'int'> | 15 | threshold for active speaker detection. Higher values are stricter. |
active_speaker_flag | <class 'str'> | 'active_speaker_flag' | field name to store the active speaker flags. |
args | '' | extra args | |
kwargs | '' | extra args |