Files
icassp2027-experiment/literature_notes/17_action_sensitivity_learning.md
2026-08-26 17:18:50 +08:00

15 lines
914 B
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.
# Action Sensitivity Learning for Temporal Action Localization
- 会议:ICCV 2023
- 阅读深度:摘要级筛选
- 官方链接:https://openaccess.thecvf.com/content/ICCV2023/html/Shao_Action_Sensitivity_Learning_for_Temporal_Action_Localization_ICCV_2023_paper.html
## 摘要(中文转述)
Action Sensitivity Learning 针对时间动作定位中不同类别、不同实例对边界和上下文的敏感程度不同的问题,通过类别级和实例级敏感性建模调整梯度贡献及对比采样,使模型更关注难定位、边界模糊的样本。
## 与当前项目的关系
它提供了比 RL 更低风险的 transition 强化路径:根据距 onset/offset 的距离、转换类型和样本不确定度重加权,但仍保持标准监督学习。建议先与均匀采样做严格对照,并保证权重归一化,避免仅因提高 transition 占比而破坏整体概率校准。