Performance comparison of CNN, QNN and BNN deep neural networks for real-time object detection using ZYNQ FPGA node. (January 2022)

• Record Type: Journal Article
• Authors: Mani, V.R.S.; Saravanaselvan, A.; Arumugam, N.
• Is Part Of: Microelectronics journal. Volume 119(2021)
• Publication Date: 2022-01
• Subjects: Deep neural networks -- Real-time object detection -- Field programmable gate array -- Convolutional neural network -- Binarized neural network -- Quantum neural network; Microelectronics -- Periodicals; Microélectronique -- Périodiques; Microelectronics; Electronic journals; Journals - contents and abstracts; Periodicals; 621.3805
• DOI: 10.1016/j.mejo.2021.105319
• Languages: English
• ISSNs: 0959-8324

Abstract

Abstract: In this manuscript, previously trained Convolutional neural network (CNN), Quantum Neural Network (QNN), and Binarized Neural Network (BNN) models performed employing Tensor Flow's Application Programming Interface (API) for real-time object detection and implemented on FPGA. Then, the proposed real time objects detection based on CNN, QNN and BNN Deep Neural Networks classifier mode activated on python, and then the dataset taken from PASCAL VOC. For an accuracy analysis of real time objection detection, this real time objects detection based on CNN Deep Neural Networks classifier provide 3.458% and 1.600% higher accuracy value than proposed real time objects detection. Then, the proposed real time objects detection based on CNN, QNN and BNN Deep Neural Networks classifier model verified by using the Verilog programming language in the Xilinx ISE 14.5 design tools in the ZYNQ FPGA development team. These results show the FPGA implementation of this real time objects detection based on CNN Deep Neural Networks classifier model meets the objective efficiently.