Big Data Multimedia Mining: Feature Extraction Facing Volume, Velocity, and Variety

A modern multimedia mining system needs to be able to handle large databases with varying formats at extreme speeds. These three attributes, volume, velocity and variety, together define big data primarily. This chapter presents the latest original research results of a showcase big data multimedia mining task by evaluating the pretrained convolutional neural network-based feature extraction through process parallelization, providing insight into the effectiveness and high capability of the proposed approach. It discusses the common strategies adopted to make data-mining scalable in terms of volume and velocity, when the variety of the data has been duly considered that is when the framework to represent the data in a consistent form is in place just as necessary. The chapter discusses “scalability through feature engineering”, which is just the process of intelligently picking the most relevant features going by the data modality and common queries.