multimodal sequential modeling and recognition of human activities published presentations and documents on DocSlides.