Design a processing method to reduce hardware resources on the Deep Learning Control server

Abstract

This system design method is proposed to reduce the use of hardware resources in systems with increasing usage of artificial intelligence. Typically, in the use of deep learning to process images, one learned model is called up and loaded onto the hardware to be used to perform the processing process. However, the resource use of one model is large, and there is a problem that occupies the process for data processing. This makes it difficult to process a large amount of data at once and what is being done in real time. In this paper, we propose a method for processing multiple data with one hardware by designing optimal tasks for processing according to the size of the image data being entered.