Optimized GPU Acceleration Algorithm of Convolutional Neural Networks for Target Detection
Shijie Li, Yong Dou, Qi Lv, Qiang Wang, Xin Niu, Ke Yang · 2016
Target detection is a hard real-time task for video and image processing. This task has recently been accomplished through the feedforward process of convolutional neural net-works (CNN), which is usually accelerated by general-purpose graphic units (GPUs). However, there is a challenge for this task. The running speed remains to be improved. In this paper, we present an efficient image combination algorithm to accelerate the feedforward process of CNN. We run our experiments on a GTX980 card. Compared with that of cuDNNv3(a standard Nvidia deep learning library), a high speedup of 6.97x is obtained in the detection task.