Les commentaires sont clos.

18-months postdoc in deep learning for multimodal underwater imaging data at LIFO, Orleans

3 Mai 2022

Catégorie : Post-doctorant

Postdoctoral position or Research Engineer - University of Orléans, France

Subject: Weakly supervised learning for multimodal underwater imaging data

Keywords: multimodal, weakly supervised learning, self supervised learning, object detection and classification

Place: LIFO - Laboratoire d'Informatique Fondamentale d'Orléans, France

Starting: 01/09/2022 (approximative). Length: 18 months.

Criteria of eligibility: have a PhD

Salaire (en fonction de l'expérience) : 2000€ à 2200€ net.

Link :


Context of the project

This project is interested in working on technological challenges in the development and deployment of deep models on the platform of an ROV (Remotely Operated underwater Vehicle) working in an underwater environment. The objective of this project is to bring innovations in auto/weakly supervised learning and in the design of efficient deep models. These make it possible to overcome the challenges of designing an intelligent ROV for recognition taches in an underwater environment. New databases for the detection and classification of specific underwater objects (mines, fish) are also built within the framework of this project to facilitate the learning of deep models as well as evaluation of various existing methods


Description of the work

The present work focuses on the study of self-supervised methods to learn the semantic/structural representation of the seabed and on weakly supervised learning to build object detectors. To do this, we will use the information already presented in the data without the need for manual labels, or with limited labels.

  • RGB Images/Vidéos
  • Sound Images

Solutions to be investigated include, but are not limited to: image transformation techniques to generate automatic labels; pretext tasks research and deep architecture design in order to train models with limited or zero labels.

We wish to propose an approach to learn the presentation of the seabed and underwater environments which is beneficial for many applications, not only object detection/segmentation (fish, mines) but also anomalous event detection, object tracking, etc. The position includes also the below tasks:

  • Management of heterogeneous training data and labels
  • Designing and evaluation of deep learning models (CNNs, Transformers, Autoencoders, etc.)
  • Periodically present the progress to the group
  • Writing of scientific reports and articles
  • Collaborate with the team members via discussions, study groups, guiding students/interns/research officers.


Job profile:

The candidate is requested to have a PhD, with a strong background in Computer Science/Mathematics/Statistics/Computer Vision or relevant fields. An important knowledge of Python is primordial. This position requests knowledge and methodologies in machine learning, computer vision.

Students expecting to finalise their degree in the coming months are also welcome to apply.


To apply or for further information please contact:

Your CV with names of two references has to be sent to the following e-mail:

Vincent NGUYEN (