An automated approach to retrieve lecture videos using context based semantic features and deep learning
- PDF / 1,251,051 Bytes
- 11 Pages / 595.276 x 790.866 pts Page_size
- 46 Downloads / 218 Views
Indian Academy of Sciences Sadhana(0123456789().,-volV)FT3 ](0123456789().,-volV)
An automated approach to retrieve lecture videos using context based semantic features and deep learning N POORNIMA1,2
and B SALEENA1,*
1
School of Computer Science and Engineering, Vellore Institute of Technology, Chennai, India School of Computing, SRM Institute of Science and Technology, Chennai, India e-mail: [email protected]; [email protected]
2
MS received 17 February 2020; revised 9 July 2020; accepted 17 August 2020 Abstract. Video digitization is one of the emerging technologies holding significant importance over years in applications like video recording and video compression. There are different techniques available in the literature for the effective retrieval of videos. This research work presents a video retrieval scheme based on a deep learning strategy. Initially, the video archive is subjected to the keyframe extraction, for extracting useful keyframes from the video. The features extracted from the keyframes are stored in the feature database. The features are clustered using the Fuzzy C Means (FCM) algorithm. These clustered features have been provided to the deep learner for finding the optimal centroid for the incoming user query. For experimentation, the research has considered videos from different categories, and both the text query and the video query have been used for the retrieval. The experimental results demonstrate the efficiency of the proposed deep learning strategy in video retrieval and its achievement of improved values of 0.9620, 0.9682, and 0.9652 respectively for recall, precision, and F-measure. Keywords.
Video retrieval; keyframes; clustering; deep learning.
1. Introduction Digitization of the lecture contents by recording it in the form of videos has helped universities and colleges in the improvement of teaching skills [1]. Students prefer learning materials that are in the video format as they are easily available on online platforms. Thus, the lecture videos have improved the robustness of the study material [2]. Video is the combination of the text, image, and sound, and hence, the use of the lecture videos as the study material enables a live study experience for the students [3]. Also, universities can upload the study material in their portals to make it readily available for the students. Using lecture videos as study material has been in vogue in recent years in most of the universities. Some colleges record the presentation of the lecturer and upload it in the internet platform [4]. Direct recording of the presentations may increase the multimedia content on the internet and it is extremely difficult for the students to find the actual content from a large source of information [1]. The ever increasing demand for lecture videos has given rise to a video retrieval system that analyzes the large database and retrieves similar video content related to the query issued by the user. As the video archives on the internet are very large, retrieval of similar video contents
Data Loading...
