Analyzing the behavior of humans in continuous video recordings is still a very difficult task. In the fully supervised setting, temporal models like RNNs are trained on videos that are annotated at a frame-level. Acquiring such annotations, however, is very time consuming and strong temporal models require large amounts of annotated training data. Weaker forms of supervision like transcripts are therefore investigated to learn temporal models. In this talk, I will describe some of our recent works on weakly supervised learning of actions and I will give an overview of the research activities that are conducted within the DFG research unit "Anticipating Human Behavior" at the University of Bonn.
This video may be embedded in other websites. You must copy the embeding code and paste it in the desired location in the HTML text of a Web page. Please always include the source and point it to lecture2go!
