Open AccessA Hardware Accelerator Based on Neural Network for Object Detection
Author(s) -
Tong Zhao,
Li Qiao,
Qinghua Chen,
Qingsong Zhang,
Na Li
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1486/2/022045
Subject(s) - computer science , frame rate , field programmable gate array , artificial neural network , titan (rocket family) , central processing unit , object detection , hardware acceleration , frame (networking) , deep learning , energy consumption , real time computing , computer hardware , artificial intelligence , embedded system , pattern recognition (psychology) , engineering , telecommunications , electrical engineering , aerospace engineering
At present, deep learning algorithms such as neural networks are widely used in all aspects of artificial intelligence. The computing performance of the CPU is low, and the power consumption of the GPU is large. This topic uses FPGA to study and implement the target detection algorithm. The proposed related technology can meet the accuracy requirements of the unmanned system for target detection, as well as the real-time requirements of the video stream. The frame rate of this design is 23.1FPS, and the mAP value is basically the same as that of CPU and GPU. The delay is 47.2% of Intel CPU, and the energy efficiency is 26% of Titan X GPU.