Zero-shot skeleton-based action recognition

Zero-shot skeleton-based action recognition is a computer vision technique aimed at recognizing action categories that have not appeared in the training set through human skeletal data. This technology leverages cross-modal transfer learning and semantic embedding to achieve accurate recognition and classification of new actions without the need for additional labeled data. Its goal is to enhance the generalization and adaptability of action recognition systems, expanding their application scenarios. It has significant application value in fields such as intelligent surveillance, human-computer interaction, and virtual reality.