Crowdsourcing Evaluation of Video Summarization Algorithm

technique of video summarizing involves selecting the most relevant and informative sections of a video to generate its shortened and faster version. Crowdsourcing is a relatively new term that has been exploited in the present study to achieve video summarization. This technique helps in dividing a task into multiple parts, and each of these parts is then evaluated using a large group of individuals to solve problems that are otherwise difficult to solve using traditional computational machines. In this study, we offer a crowdsourcing subjective experiment in which summaries of processed video sequences are evaluated. Thus, we are proposing an experiment that utilizes crowdsourcing to evaluate the efficacy of an algorithm that summarizes videos. A group of 45 individuals participated in the experiment, where each of them were asked to watch 24 videos, each of 30-second and 45-second duration. An experimental comparison was conducted with respect to presentation order and random selection methods. A content-based video segmentation was also used to represent different levels of complexities and visual richness. The findings of the assessment showed that specific characteristics of a video such as its length, complexity, and content, play a major role in improving the performance of the summarization algorithm. This study is an essential step toward the development of video summarizing systems that are both more accurate and more efficient.