Work Description: Join us in pioneering the future of 3D scene representation and rendering
with Neural Radiance Fields (NeRF). Initially, NeRF models were resource-intensive to train,
but recent advancements have significantly improved their efficiency. In the automotive
industry, rapid and efficient inference is crucial. This project aims to push the boundaries of
embedded Neural Networks by exploring their potential in new applicative domains.
Your mission will be to optimize Neural Rendering models to render the vehicle’s surrounding
environment in real-time. We will place a special emphasis on the latest developments in NeRF
technology, striving for a real-time architecture. Typically, speeding up NeRFs involves a tradeoff
with rendering precision. However, innovative approaches, such as dividing scenes into
thousands of voxel grids with independent small NeRFs, are being explored to maintain visual
quality while enhancing speed.
Key Responsibilities:
– Develop and optimize Neural Rendering models for real-time applications.
– Explore and implement efficient training methods on embedded hardware.
– Collaborate with a dynamic team of researchers and industry experts.
– Contribute to the advancement of state-of-the-art 3D scene representation
technologies.
– Specific application domain to be discussed face-to-face.
What We Offer:
– A stimulating research environment at Mines Paris PSL University.
– Collaboration with Forvia, a leader in the automotive industry.
– Opportunities to publish and present your work at leading conferences.
– A chance to make a significant impact on the future of automotive technology.
Qualifications:
– PhD in Computer Science, Electrical Engineering, or a related field.
– Strong background in Neural Networks and 3D rendering.
– Experience with embedded systems is a plus.
– Passion for innovation and problem-solving.
Join us in this exciting journey to revolutionize automotive technology with cutting-edge Neural
Rendering solutions. Apply now and be a part of the future !
References:
– Representing Scenes as Neural Radiance Fields for View Synthesis, Tancik et al., 2020
– Instant Neural Graphics Primitives with a Multiresolution Hash Encoding, Müller et al.,
2022
– Neural Radiance Fields for Unconstrained Photo Collections (NeRF in the wild), Brualla
et al., 2021
– FastNeRF: High-Fidelity Neural Rendering at 200FPS, Garbin, Stephan J et al, arXiv
preprint 2021
– PlenOctrees for Real-time Rendering of Neural Radiance Fields, Alex Yu et al, 2021,
ICCV 2021
– Baking Neural Radiance Fields for Real-Time View Synthesis, Peter Hedman et al,
ICCV 2021, https://nerf.live
– KiloNeRF: Speeding up Neural Radiance Fields with Thousands of Tiny MLPs,
Christian Reiser, ICCV 2021
Contacts:
Bogdan Stanciulescu: MINES Paris PSL: bogdan.stanciulescu@minesparis.psl.eu
Samer Ammoun: samer.ammoun@forvia.com
