Data Center Monitoring Using an Improved Faster Regional Convolutional Neural Network
Ankit Narendrakumar Soni · SSRN Electronic Journal · 2018
The Data Center contains loads of servers whose pointer LEDs can give the issue data, which is significant for data security. A novel server recognition conspires joined with deep learning, and recognition computing was proposed to screen the server's functional status progressively. In this technique, the best in class Faster RCNN system was improved by fitting anchors selection, hard negative mining, and non-greatest concealment. Morphological tasks were utilized to fortify the vigor of the customary LEDs location calculations. For the Resnet model, our framework accomplished an edge pace of 14 fps and article exactness of 96% on an NVIDIA Titan X. The proposed algorithm got excellent location execution in natural conditions, making it considerably more exact and proficient at screening the servers' deficiency data.