ResearchBib Share Your Research, Maximize Your Social Impacts
Sign for Notice Everyday Sign up >> Login

2-year post doc position in deep learning at Saint-Etienne

Country : France

Website : http://univ-st-etienne.fr

Description

Semi-supervised anomaly detection in ski lift videos via active learning
Context
Safety in ski lifts is a major concern for ski resort operators. To prevent possible accidents, it is necessary to detect dangerous
situations on ski lifts as early (after boarding) as possible.
The main goal of the MIVAO project is to develop a machine learning and computer vision system able to do real-time analysis of videos
acquired by a camera fixed on one of the first pillar of the chair lift,
and trigger an alarm when a problem (safety bar not in place, child
alone on the chair, two many persons, child sliding under the safety
bar, etc.) is detected. This cutting edge project is led by the BlueCime
company (located in Grenoble, France) in partnership with the Hubert
Curien Laboratory (Saint-Etienne, France), the Gipsa Lab (Grenoble) and an increasing number of ski resorts in the Alpes mountains.
The Post-doc candidate will join the working team of Mivao composed ofthree PhD students and more than 10 researchers and engineers.
The candidate will be hosted in the Hubert Curien Laboratory in
Saint-Etienne, working with and contributing to the supervision of 2 Ph.D. students.
Scientific objectives
The objective of this project is to propose new machine learning methods
to analyze the acquired videos and trigger an alarm when an anomaly isdetected.
These methods will be based on deep convolutional networks which have shown outstanding performance on related applications but
also on this specific dataset, in a supervised setting. Because the
system can continuously acquire videos and because labeling those videos
is expensive, the post doc will mainly focus on unsupervised,
semi-supervised and active learning methods. The main issues to addresswill be:
1. Study, implement and test unsupervised methods in deep learning for anomaly detection
2. Propose new unsupervised and semi-supervised methods for anomaly
detection
3. Propose new active learning methods with domain adaptation, to
benefit from a few well-labeled examples to improve an
existing system.
Required skills
We are searching for an outstanding and highly motivated candidate with:
- A PhD degree in machine learning with preferably some knowledge in computer vision and deep learning.
- A list of excellent publications (at reference journals and
conferences in machine learning, pattern mining and/or
computer vision) that demonstrates the expertise of the candidate
- Very strong programming skills in languages such as Python (and,
ideally, an experience in Tensorflow). The candidate will need to
show during the interview, examples of codes that
he/she implemented.
Working environment
Saint-Etienne is a mid-size city (14th biggest city in France) and oneof the cheapest in terms of living cost (accommodation, food).
It is 70km from Lyon (45 min by train) and in a middle of a splendid regional park where skiing, hiking, climbing, biking are possible. The hosting
research group has established expertise in relevant domains including statistical machine learning, transfer learning, unsupervised learningand computer vision. `
The successful candidate will have the opportunityto work in synergy with two Ph.D. students.
The **expected salary is 2193 euros NET per month** (gross salary: 2727 euros).
Application instructions
The application consists of a motivation letter, CV (with a detailed list of publications
and links to e.g implementations on Github), names and contact details of two references.
The post-doc should ideally start before the end of 2017. Applications
will be reviewed continuously and an answer will be sent within a month after the
application date.
Contact
Elisa Fromont,
(mailto:elisa.fromont-AT-univ-st-etienne.fr)
Phone: (+33) 4 77 91 57 67
Rémi Emonet,
mailto:remi.emonet-AT-univ-st-etienne.fr
Phone: (+33) 4 77 91 57 23

Last modified: 2017-07-30 10:17:45