Big Data growth does not stop. Companies are forced to store data without knowing its value or usage because data has potential to become next revenue opportunity. Since the cost for storing data increases as the amount of data grows, efficient data storage methods are required. We introduce a method to reduce the cost of data storage by partial decluttering of image and video data. In our method, only the portions of image or video files containing recognized objects are extracted. This is based on the assumption that images of objects of interest may be used in the future, but other image data, such as the background, can be discarded. We use the neural network computer stick from Intel Corporation to accelerate the computation. In the case study of a video file, our method successfully reduced the data size by approximately 50% on average. In addition to that, we realized a low-cost method that can be applied not only to large companies but also small to medium companies by using a low cost hardware accelerated implementation. We interviewed experts in corporate IT data management and qualitatively demonstrated that our method is valuable and effective. Our research has contributed to IT managements' decision-making support to resolve their data storage vs. cost problem.
展开▼