About this document
Unsupervised Human Action Recognition by dasepsilon is a document available to read on EtoBox.
This document presents a novel unsupervised learning method for human action recognition in videos. The method represents video sequences as collections of spatial-temporal words by extracting space-time interest points. A probabilistic Latent Semantic Analysis (pLSA) model is used to automatically learn probability distributions of the words and action categories from unlabeled video data. The learned model can then categorize and localize human actions in new video sequences.
- Author
- dasepsilon
- Language
- EN