LSTM for Action Recognition

Sofia is attending Massachusetts Institute of Technology majoring in Computation and Cognition and minoring in Linguistics. Her goal is to innovate novel approaches in computer vision and natural language processing for application in the next generation of Robotics. A long short-term memory (LSTM) network is a specialized form of recurrent neural network. These networks use feedback connections which allow for the processing of sequences of data as opposed to single points, and thus are capable of learning long-term dependencies. For the purpose of action recognition with an LSTM, we may represent each action as a series of single poses, our atomic unit.