SUMMARIZATION OF CAPTIONS GENERATED FOR A VIDEO THROUGH LSTM AND GRU USING TF-IDF VECTORIZATION
- Preethi1, Dr. P. Dhanalakshmi2and Dr. T. Thiruvengatanadhan3
1Research Scholar, Department of Computer Science and Engineering, Annamalai University, Tamilnadu, India
2Professor, Department of Computer Science and Engineering, Annamalai University, Tamilnadu, India
3Assistant Professor, Department of Computer Science and Engineering, Annamalai University, Tamilnadu, India
Video data is very much helpful in easy understanding of a particular event. Nowadays, the video lectures play a major role among the students to have a better understanding of the concepts. Cooking videosin YouTube also found to be more interesting among women. But when these kinds of videos are too lengthy, the people won’t watch the video fully. So, if the video is summarized, then the people may have better understanding and save their time in watching the full video. Summarization helps to save the storage space and time. In this work, video shot boundary detection is performed to identify the number of shots in the lengthy video. After that, the captions are generated for the shots of video using video captioning technique. Finally, the dense captions generated are summarized using TF-IDF method. Thus, the long video could be summarized with the help of very few lines of natural language sentences.
Keywords – TF-IDF, Caption, Shots