Files
icassp2027-experiment/literature_notes/34_sicnet_addressee.md
2026-08-30 15:36:28 +08:00

14 lines
1.2 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# Long-Term Social Interaction Context: The Key to Egocentric Addressee Detection
- 作者:Deqian Kong、Furqan Khan、Xu Zhang、Prateek Singhal、Ying Nian Wu
- 发表:ICASSP 2024
- 阅读深度:重点方法与输入设定;未运行
- 记录日期:2026-08-28
- 来源:[Amazon Science](https://www.amazon.science/publications/long-term-social-interaction-context-the-key-to-egocentric-addressee-detection)、[论文](https://cdn.amazon.science/64/e3/22da99d74e3097774c7fa907374f/long-term-social-interaction-context-the-key-to-egocentric-addressee-detection.pdf)
## 摘要与项目关系
SICNet 在 Ego4D talking-to-me 中结合较长对话历史和人物视觉上下文。对话分支编码转写及 ego/non-ego 信息,视觉分支编码人脸、身体与位置,再融合做讲话对象分类。
启发是单纯“谁在讲话”不足以确定 wearer 是否应接话。限制是它在话语结束处分类,并使用数据集 diarization 和提供的框;不能把结果直接当作无真值输入的连续预测。我们应将讲话对象作为训练辅助目标,输入仍保持因果混合 AV。官方 TTM 划分与 EgoSpeak 划分需要独立核对。