Internships at Unity Labs Grenoble

Interested in Deep Learning and/or Computer Graphics? Want to do a research internship, using and/or extending state-of-the-art techniques? Then we should get in touch. Whether you know already what you'd like to work on or want to collaborate directly on my research topics: I'm always very happy to mentor research internships. At Unity Labs Grenoble we work on cutting-edge research in the field of Computer Graphics (real-time and offline rendering) and (semi-)automated content creation leveraging (deep) machine learning. We provide a stimulating, friendly and comfortable work environment. Below, you'll find some possible internship topics.

Interested? Drop me an e-mail (in either French or English) describing in what context (e.g., studies) you want to do your internship, what your interests are, what your knowledge is in the field, when you'd probably want to do your internship, and any questions you may have.

Requirements:

Be at least at master/graduate level in your studies. Studies should encompass one or more of the following fields: computer graphics, computer vision, machine learning.
Have some hands-on experience with deep learning tools (preferably but not necessarily TensorFlow)
Experience with Computer Graphics/rendering projects is a plus.
Development skills with the Unity Editor is a plus.
A visible portfolio of projects (e.g., on GitHub or personal website) is a big plus.

By-example content creation

We wish to provide artists with tools to reproduce appearances photographed in the real world:

Example-based texture/material synthesis and style transfer

[Research and/or development]

Given exemplar images of material and/or texture, can we (semi-)automatically populate a virtual world with a similar appearance? This requires to be able to copy the exemplar material, but with variation (to avoid repetition artefacts) and without visible seams. Neural texture synthesis is a promising tool for solving this problem. Several internship subjects are available in this line of work, ranging from very research-oriented (e.g., the final goal would be to write a research paper) to pure engineering tasks (e.g., the final goal would be a Unity plugin).

Differentiable Rendering

[Research internship]

A rendering engine like Unity calculates the rendered image given input parameters (the scene's geometry, materials and light sources) using global illumination. A long-standing goal is to be able to reverse the process: given a rendered or photographed image, can we infer the input scene's parameters. Solving this ill-posed problem would allow to edit these parameters, thus modify such a scene (e.g., relighting, object insertion, material editing). Differentiable rendering allows to express the forward problem (global illumination) in a differentiable manner. Hence one can backpropagate gradients of errors through this process using stochastic gradient descent (SGD) thus optimize for variable input parameters. Several research internship opportunities can be imagined on this topic, in collaboration with my "rendering" colleagues at Unity Labs Grenoble.

Rendering post-processing using neural networks

[Research and development internship]

The most recent NVidia GPUs (RTX Turing) offer capabilities for real-time image post-processing using neural networks: e.g., denoising, superresolution, etc. Soon, many video games will incorporate such capabilities to improve frame-rates. Internships in this field involve experimenting with these capabilities in the context of the Unity Engine.

Previously supervized students and PhD candidates

Note: old or future projects you see below have not been translated, they are in their initial language (French or English).

ETH Zürich

PhD candidates

Nikolay Kobyshev, ETH Zürich
Louis Lettry, ETH Zürich

Master theses (6 months)

Tijmen Verhulsdonck (Master 2 student in "Embedded Systems", KTH Royal Institute of Technology, March - August 2017): One Shot Object Detection: For Tracking Purposes.
Andrey Ihnatov (Master 2 student in "Computer Science", ETH Zürich, November 2016 - April 2017): Photo Quality Enhancement with Deep Convolutional Neural Networks.

Semester theses (Master level, 4-8 weeks)

Andreea Lutac (Master 1 student in "Robotics", ETH Zürich, Spring 2017 - 12 weeks): Creating an annotated texture database for automatic texture model inferences from exemplars.
Dario Fuoli (Master 1 student in "Information Technology and Electrical Engineering", ETH Zürich, Fall 2016 - 7 weeks): Summarizing façade textures through texture synthesis: some deep learning tests.
Samuel Keusch (Master 2 student in "Computer Science and Engineering", ETH Zürich, summer 2016 - 160h): Where is this webcam located? Preliminary implementation.

Inria

Free internship

Georgios Kopanas (Master 2 student from University of Thessaly, Department of Electrical and Computer Engineering, autumn 2015 - 3 months): CUDA parallelization of an algorithm for procedural texturing.