Identifying Head Movements in 360- Degree Videos - A Convolutional Neural Network Approach

Panoramic videos being an interactive media are gaining a lot of popularity. They give a horizontal elongated view where the Field of View (FoV) in the range of 360o × 180o. FoVs are important to capture the material to be included in the video sequences. The part outside FoV range is not included in the video sequences. This is controlled by human head movements (HM) performed by wearing the head mount displays. HM prediction is at extreme level of importance and thus needs attention as it determines the eye fixations too. Thus, head movements and eye fixations are a vital part for perfect capturing of the material in the FoV range. There are works related to this comprising of the deep learning approaches, reinforcement learning and more. The paper establishes a database with the help of face detection and camera. Collecting this database, we will use the convolutional neural networking approach to capture the environment in the FoV.