Optimized Approach for Video Summarization using Transfer Learning and LSTM
Abhishek Dhiman, Maroti Deshmukh · 2023
Video content has been increasing at a very high pace, so the need for summarizing videos is of urgent need. Video summarization emphasizes on quick go-through of video content. In the last few decades, the field of video summarization has inspired a lot of research. Many of the techniques have been proposed by different researchers, but most of them are not able to create summaries of general videos, i.e., they are generally suitable for a set of categories of videos. To overcome this problem, we have introduced a general framework for video summarization, where we have used a pre-trained VGG16 model for feature extraction, and based on the features, we found the frame level importance using a Long Short Term memory (LSTM) network. After that, we selected the frames with high frame-level scores and finally combined them to form the video summary. Experimenting with the different types of videos taken from the different datasets shows that the proposed methodology outperforms the existing state-of-the-art methods.