By continuing to browse the site, you are agreeing to our use of cookies. (More details)

PhD offer "Ultra-fast vision using Spiking Neural Networks"

This offer is available in the following languages:
Français - Anglais

Ensure that your candidate profile is correct before applying. Your profile information will be added to the details for each application. In order to increase your visibility on our Careers Portal and allow employers to see your candidate profile, you can upload your CV to our CV library in one click!

Faites connaître cette offre !

General information

Reference : UMR7289-LAUPER-001
Workplace : MARSEILLE 05
Date of publication : Tuesday, June 30, 2020
Scientific Responsible name : Laurent PERRINET
Type of Contract : PhD Student contract / Thesis offer
Contract Period : 36 months
Start date of the thesis : 1 October 2020
Proportion of work : Full time
Remuneration : 2 135,00 € gross monthly

Description of the thesis topic

Biological vision is surprisingly efficient. To take advantage of this efficiency, Deep learning and convolutional neural networks (CNNs) have recently produced great advances in artificial computer vision. However, these algorithms now face multiple challenges: learned architectures are often not interpretable, disproportionally energy greedy, and often lack the integration of contextual information that seems optimized in biological vision and human perception. Crucially, given an equal constraint on energy consumption, these algorithms are relatively slow compared to biological vision. It is believed that one major factor of this rapidity is the fact that visual information is represented by short pulses (spikes) at analog – not discrete – times ( Paugam and Bohte, 2012). However, most classical computer vision algorithms rely on such frame-based approaches. One solution to overcome their limitations is to use event-based representations, but these still lack in practice, and their high potential is largely underexploited. Inspired by biology, the project addresses the scientific question of developing a low-power sensing architecture for the processing of visual scenes, able to function on analog devices without a central clock and aimed at being validated in real-life situations. More specifically, the project will develop new paradigms for biologically inspired computer vision ( Cristobal, Keil and Perrinet, 2015), from sensing to processing, in order to help machines such as Unmanned Autonomous Vehicles (UAV), autonomous vehicles, or robots gain high-level understanding from visual scenes.
More info @ https://laurentperrinet.github.io/post/2020-06-30_phd-position/

Work Context

The thesis will be carried out in the team “NEuronal OPerations in visual TOpographic maps” (NeOpTo) within the Institut de Neurosciences de la Timone in Marseille, a lively town by the Mediterranean sea in the south of France. The research team is led by F. Chavane (DR2, CNRS) and currently hosts 4 permanent staff, 3 post-docs and 4 PhD students. The research themes of the team are focused on neuronal operations within visual cortical maps. Indeed, along the cortical hierarchy, low-level features such as the position and orientation of the visual stimulus (but also auditory tone, somatosensory touch, etc…) but also higher-level features (such as faces, viewpoints of objects, etc…) are represented topographically on the cortical surface.

This work will be conducted in direct collaboration with Jean Martinet. We will develop these algorithms in collaboration with Ryad Benosman (Université Pierre et Marie Curie) and Stéphane Viollet (équipe biorobotique, Institut des Sciences du Mouvement).

Constraints and risks

IT development work that does not present any particular constraint or risk.

Additional Information

This thesis is funded by the European project https://www.chistera.eu/projects/aprovis3d and will require a pro-active collaboration with the consortium partners (France, Switzerland, Spain, Greece).

We talk about it on Twitter!