Mistake Detection

Error detection is a sub-task in the field of computer vision that aims to classify each rough action segment by modeling process knowledge and retaining long-distance sequence information, identifying whether it belongs to one of three categories: correctly performed, erroneous operation, or corrective behavior. The goal of this task is to improve the accuracy and timeliness of AR assistants in aiding users to perform complex tasks, which has significant application value and can effectively enhance user experience and work efficiency.