Optimisation of an industrial image processing code for defect detection

Description de l'offre

Country:

FRANCE

Region:

LORRAINE

Sector:

ArcelorMittal Europe - Flat Carbon

Unit:

ARCELOR RESEARCH

Site:

ArcelorMittal RESEARCH Maizieres

Field of activity:

Research and Development / Metallurgy Innovation

Title of internship:

Optimisation of an industrial image processing code for defect detection

Subject of the internship:

The internship consists into applying state of the art neural networks for defect detection in images and testing GPU configuration for speed improvement The mission : accountabilities and activities In a previous three year PhD, ArcelorMittal Maizieres Research has developed tools based on convolutional neural network and deep learning to process images with the final goal of applying these algorithms to defect detection and classification on steel surfaces. These algorithms are now functional, but have to be instantiated, tuned, and maybe improved to go from theoretical research to effective industrial solutions. The successful candidate will be in charge of applying these algorithms to real industrial image data sets, with the goal of finding the best possible working configuration and evaluating the performances that can be obtained for steel surface defect detection. One of the draw-backs of such algorithms is the computation time that can be intensive at learning stage. It might be interesting to try to set-up a working GPU configuration to reduce the training time. The environment The internship is based in Maizieres-Lès-Metz within ArcelorMittal biggest research facilities (>500 persons). The trainee will be integrated in the Measurement and control department in the Data and signal processing team. This is a highly multicultural environment with engineers and trainees from many countries in the world. The training will be supervised by image processing engineers and will be working with state of the art tools in the field of neural network and deep learning.

Profil recherché

Applicant Profile:

The successful candidate should have very good programming and scientific skills. The following competencies are required: • Good level in mathematics and machine learning is required. • Python language is likely to be used for tests purposes on indus

Education:

Computer sciences

Level of education:

MASTER (Bac+4/5)

Language:

French - English

Duration of the internship:

6 months

Period of the internship:

MARCH 2016

Managers of the internship:

Sophie BAGNASCO
Tel:
Email: ************@*******.**

Comments:

Envoyer vos CV et lettres de motivation en indiquant la référence du stage a l'adresse suivante : ************@*******.**

© 2007 -