Procedure Step Recognition

Procedure Step Recognition (PSR) is a subtask in the field of computer vision that focuses on recognizing the correct completion and order of procedural steps. The goal of PSR is to extract all the steps correctly performed by a person up to time $t$, based on sensor inputs $X_t=(x_t, x_{t-1}, \dots, x_{t-h})$ at time $t$ and the set of procedural actions to be executed $\mathcal{P}=\{a_0, a_1, \dots, a_n\}$. Unlike traditional action recognition, which may focus on partial execution of actions, PSR emphasizes the completion status and correctness of procedural steps. This provides a more meaningful perspective for understanding procedural videos and has significant application value.