The Xerox Research Centre Europe (XRCE) has a strong expertise in Image Understanding, applied with success in domains such as parking automation or traffic rules enforcement. Automating such image analysis is obtained by the use of machine learning algorithms, with the need of labelled images to construct the necessary models.

The internship proposal aims at designing and developing a process to learn a recognition model for a new object from a 3D model of the object. Using synthetic data in machine learning has several advantages: it reduces the task of manual annotations and a large amount of images can be created through the simulation of the real world environment (illumination changes, points of view…). The major goal of this project is to compare the performances of recognition models trained on synthetic data from models trained on real world data, as well as evaluating the amount of work required to train a new object.

In this internship, the intern will focus on some of the following tasks in collaboration with the other team members:

  • Evaluate 3D scanning technologies and design simple 3D models of objects
  • Extract 2D images after transformations (lights, points of view) and create a database of synthetic images
  • Generate the input parameters for the training system
  • Evaluate the performances of the generated model against a model trained on real world data.

The intern will be part of a team of developers and researchers. XRCE projects are managed with an agile methodology which allow interns to follow all the steps of software development, including research, design, implementation, test and integration.

Candidates will be:

  • Students in Computer Sciences Engineering\Bsc\Masters
  • Strong knowledge AND practice of Object Oriented Programming: C++ and Java in particular
  • Knowledge in 3D engines, 3D rendering and simulation
  • Knowledge in computer vision and machine learning techniques
  • Knowledge in Agile methodologies and Test Driven Development is a plus

During her/his internship the candidate will acquire a significant knowledge in advanced computer vision and machine learning techniques while working closely with researchers and engineers. Additionally, the candidate will become knowledgeable of Agile methodologies (eg: SCRUM, Continuous Integration)

Start Date:
Q1 2016
4-6 months
Apply instructions:

To submit an application, please send your CV and cover letter to both  and Please specify "Internship: Design and evaluation of a process for training categorization models from synthetic data" in your subject line.