Finding Interesting Frames in Deep Video Analytics: a Top-K Approach

摘要: Recently, the impressive accuracy of deep neural networks (DNNs) has created great demands on practical analytics over video data. Although efficient and accurate, latest analytic systems have not supported beyond simple queries like selection. In data analytics, Top-K is a very important analytical operation that enables analysts to focus most entities. this paper, we present Everest, first system supports accurate analytics. Everest ranks identifies interesting frames/moments from videos with probabilistic guarantees. built careful synthesis computer vision, machine learning, uncertain management, query processing. Evaluations five real-world Visual Road benchmark show achieves between 16.3x 20.6x higher efficiency than baseline approaches high result accuracy.