AI-Based Visual Odometry Implementation on an Embedded System


Büyüksolak O., Güneş E. O.

10th International Conference on Electrical and Electronics Engineering, ICEEE 2023, İstanbul, Turkey, 8 - 10 May 2023, pp.47-51 identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/iceee59925.2023.00016
  • City: İstanbul
  • Country: Turkey
  • Page Numbers: pp.47-51
  • Keywords: CNN, Deep Learning, Embedded System, KITTI, Visual Odometry
  • Istanbul Technical University Affiliated: Yes

Abstract

Embedded visual odometry(VO) implementation may provide a low-power, small-size alternative or compan-ion positioning system to the Global Navigation Satellite Systems(GNSS) and Inertial Navigation System(INS). As the em-bedded systems are memory scarce, in this paper, a new low-memory footprint neural network-based visual odometry method that is implementable on embedded systems is introduced and evaluated. To deploy the neural network, MAX78002 [1] artificial intelligence microcontroller has been chosen as the embedded platform. To the best of our knowledge, this is the first study that provides a microcontroller-based visual odometry solution.