PhysFormer++: Facial Video-based Physiological Measurement with SlowFast Temporal Difference Transformer
2025/7/29小于 1 分钟
What is it？
Stem + TD-MHSA + SlowFast
What is Stem？
卷积、池化、下采样，to learn video representation.
What is temporal difference multi-head self-attention (TD-MHSA)?
相邻帧减法、qkv加权求和，transformer的结构
What is SlowFast？
slow，fast两条transformer结构