Online learnable keyframe extraction in videos and its application with semantic word vector in action recognition

Video processing has become a popular research direction in computer vision due to its various applications such as video summarization, action recognition, etc. Recently, deep learning-based methods have achieved impressive results recognition. However, these need process full sequence recognize the action, even though many of frames are similar and non-essential recognizing particular action. Additionally, increase computational cost can confuse method Instead, important called keyframes not only helpful an but also reduce time each classification or other applications, e.g. summarization. As well, current yet been demonstrated online fashion. Motivated by above, we propose learnable module for keyframe extraction. This be used select key shots thus, applied The extracted input any model We plugin use semantic word vector along with novel train/test strategy models. To our best knowledge, this is first proposed. experimental on commonly datasets summarization recognition effectiveness proposed module.