Phelma Formation 2022

Project : Processing on GPU-FPGA (Image)(SICOM S9) - 5PMSPIA0

  • Number of hours

    • Lectures 0
    • Projects 32.0
    • Tutorials 0
    • Internship 0
    • Laboratory works 0
    • Written tests 0

    ECTS

    ECTS 4.0

Goal(s)

This project is interested in the implementation of deep learning algorithms applied to images on GPU processor.
The objective is to illustrate an approach of adequacy between algorithm and architecture that is necessary for an efficient implementation. Industrial flows of development are used in this teaching.

Contact Dominique HOUZET

Content(s)

The project takes place in two stages:

  • Study of the convolutional neural network with its database experimented with standard tools (Keras ...)
  • C / CUDA implementation on CPU / GPU of the different CNN layers with optimization of the implementation on GPU


Prerequisites

Test

during the last project session, there will be an oral presentation of the work carried out. A 15-20 page project report is also produced.



Rapport : 100%

Additional Information

Course list
Curriculum->Double-Diploma Engineer/Master->Semester 9
Curriculum->Engineering degree->Semester 9

Bibliography

CUDA programming guide, Nvidia