FPGA based odour recognition with TensorFlow and High-Level Synthesis

Research output: Contribution to conferencePaperpeer-review


As more and more potential applications are identified for electronic noses, the need arises to have portable platforms that can perform odour recognition in-place. Although there have been many experiments with electronic noses, the vast majority of them were done in a laboratory environment. Methods for portable e-nose systems are needed that are capable of handling the complex pattern recognition algorithms of odour recognition. In this paper we show how odour recognition can be implemented for AMD Xilinx's System-on-Chip boards. TensorFlow was used to train a convolutional neural network on two different datasets. The neural network was also implemented in C++, and Vitis HLS was used to translate that to an RTL design. We show that this workflow is a valid way to create a portable electronic nose system.
Original languageEnglish
Publication statusPublished - Dec 2022
Event2022 IEEE 1st Industrial Electronics Society Annual On-Line Conference (ONCON) -
Duration: 9 Dec 202211 Dec 2022


Conference2022 IEEE 1st Industrial Electronics Society Annual On-Line Conference (ONCON)
Abbreviated titleONCON

Cite this