Extracting Primitive Tasks from Procedural Videos using Machine Learning

Narrated instructions videos are the rich source of information. The problem of automatically extracting the procedures that are narrated in the video and to summarize the important steps to reduce the time and provide the information to the user is addressed. Natural Language Processing Technique is used for extracting the summarization of the narrated task in the given video. The input video dataset contains real world interactions between a human and machine to explain the procedure to complete a certain task. The given data set is analyzed, processed, and finally a list of important steps are extracted which will help in achieving the given task. Procedure identification or Procedure Extraction is useful for a variety of applications including intelligent tutoring systems, representation learning, transfer learning and video understanding.