We are seeking a postdoctoral researcher/research engineer to join our Smart Live Prod project, which aims to develop intelligent software solutions for the production and streaming of live video content, with a focus on content accessibility in the specific domain of public plenary broadcasts. The research engineer will primarily be responsible for researching and developing a generative model for face video compression. This model will be crucial for optimizing the efficiency of live video streaming during public plenary broadcasts, providing high-quality communication even in low bandwidth conditions.
Context:
The Smart Live Prod project aims to develop intelligent software modules to assist and automate the production and broadcasting of live video streams on the Internet, in an institutional communication context. It focuses on three main themes: automatic live video production and direction, automatic generation of accessible formats, and optimization of video compression and transport for live broadcasts of public sessions with low bandwidth. To achieve these objectives, the project relies on recent advances in artificial intelligence and deep learning technologies, particularly deep generative models.
In particular, streaming videos for public plenary sessions have specificities compared to general videos, notably containing a large quantity of close-up shots focused on speakers' faces. To code these videos more efficiently, we can rely on the very recent advances in generative coding techniques for face videos (Generative Face Video Coding, GFVC). GFVC exploits the compact representation of facial priors and the strong inference capability of deep generative models, thus enabling high-quality face video communication in ultra-low bandwidth scenarios. Recently, generative neural codecs for face videos have also become a topic of exploration in the standardization community.
Responsibilities:
- Design and develop a generative model for face video compression, building upon recent advances in the field.
- Test and evaluate the performance of the generative model in live video streaming scenarios.
- Contribute to the writing of reports and scientific papers for publication.
Required Qualifications:
- PhD in computer science, image processing, computer vision, or related fields.
- Solid experiences in deep learning.
- Proficiency in common programming languages such as Python, as well as machine learning frameworks like TensorFlow or PyTorch.
- Experience in video compression is preferred.
- Ability to work in a team and communicate research results effectively.
- Good level of written and spoken English.
- French language skills are preferred, but not mandatory.
Additional Information:
Location: INSA Rennes/IETR Lab, Rennes, France
Contract Duration: 18 months
Beginning Date: June - October 2024.
Application:
Send a CV and recommendation letter(s) to:
Lu Zhang (lu.zhang@insa-rennes.fr)
Xiaoran Jiang (xiaoran.jiang@insa-rennes.fr)
Don’t hesitate to contact us if you want to know more information about the project.
(c) GdR IASIS - CNRS - 2024.