Title: Embedded Deep Learning for Real-Time Vision Applications

Despite the great performance of deep learning methods, for applications in embedded systems, we have to consider alternative solutions to GPUs, such as custom hardware accelerators. However, current implementations are biased towards the standard evaluation pipeline of a model and limited to classification tasks with small architectures. As such, they are not evaluated in a realistic setup for manufacturers where data are streams like videos in autonomous driving. In this thesis, we target the category of applications where data are streams, real-time is required, and networks are deeper.

The candidate is expected to have (or in the process of finishing) a Master degree in Computer Science or in Electronic Engineering. A Research Master in Embedded Systems or a first experience in a research environment would be a plus.

The candidate must have experience in device design software (Quartus or Vivado) and in hardware description languages (VHDL or Verilog).

If you are interested in this position, submit a resume and a cover letter (both in English) to the following contacts.

Contacts:

  • Mohamed Amine KHELIF, ETIS lab, mohamed-amine.khelif@ensea.fr
  • Pierre JACOB, CTU Prague (Czech Republic), jacobpie@fel.cvut.cz
  • Aymeric HISTACE, ETIS lab, aymeric.histace@ensea.fr

Fichier PDF
Télécharger le PDF