Using Deep Learning REST APIs, Vid-Sum summarizes videos

Abstract

The goal of video summarizing is to condense a lengthy film into a concise one that conveys the main points and mood of the source material. This is useful since it allows us to get the main points from a shorter film without having to watch the whole thing. The majority of supervised learning video summarization algorithms now employ CNNs, with a few also using recurrent neural networks. We present VidSum, a Deep Learning architecture for video summarization. For video summarization, we use a combination of convolutional neural networks and long short-term memory (LSTM) networks. With the help of our deep learning model, we can identify which frames are most significant at certain points in time and then create video summaries that are both coherent in terms of time and include all of the relevant information from the movie. When tested on the well-known TVSum and SumMe datasets, our model achieved better results than competing algorithms when it came to video summarization.