Menu principal

Deep learning approaches for character recognition in Byzantine seal images

Information about the internship

  • Supervisors:
    • Laurence Likforman-Sulem, Attilio Fiandrotti, Département IDS, IP de Paris/Telecom Paris
    • Victoria Eyharabide, STIH Laboratory, Sorbonne Université
  • Location: Maison de la recherche, Sorbonne Université - 28 rue Serpente, 75006 Paris.
  • Duration: 12 months (starting preferably on April 2022)

Detailed topic

Byzantine seals are small circular objects (10-50 mm) used to identify the sender of letters. They carry a large part of the knowledge on the administration and the Byzantine aristocracy, but also on the cult of the saints. The seals are made of lead and have undergone alterations over time. As a result, characters may be damaged or even erased. The objective of this internship is to develop deep learning type approaches for the recognition and localization of objects and characters on seals.

We can consider deep learning architectures (FCOS, U-Nets) to propose regions of interest corresponding to the characters in the images, as well as their label (iconic object, or character). This corresponds to jointly trained end-to-end regression and classification tasks. We will use a transfer learning approach, using pre-trained networks. Then, we will develop an approach based on transformers to label a sequence of characters, from character regions.

This research will be developed within the framework of the ANR BHAI project, grant number ANR-21-CE38-0001 https://anr.fr/Project-ANR-21-CE38-0001. The candidate will work in Paris, at Sorbonne University (V. Eyharabide), in close collaboration with Telecom Paris (L. Likforman and A. Fiandrotti).

Application

Applicants should send an email to Victoria Eyharabide (maria-victoria.eyharabide@sorbonne-universite.fr) and Laurence Likforman-Sulem (Laurence.likforman@telecom-paris.fr) with:

  • A full curriculum vitae including a complete list of publications
  • A transcript of higher education records
  • A one-page research statement discussing how the candidate's background fits the proposed topic
  • Two support letters of persons that have worked with them.
Mots-clés
Computer Science / Image Processing / Document Analysis / Pattern Recognition; Computer Vision; Deep Learning-based Transfer Learning; neural networks
Établissement
Sorbonne Université
15-21 Rue de l'École de Médecine, 75006 Paris Paris  
Site Web
https://anr.fr/Projet-ANR-21-CE38-0001
Date de début souhaitée
01/04/2022
Langues obligatoires
Anglais; Français
Type de contrat
CDD
Type de poste
Postdoc
Prérequis

Applicants are required to have:
• A PhD in Computer Science.
• Advanced skills in Python programming are mandatory.
• A strong background in Machine Learning & Deep Learning on images using related libraries (scikit-learn, Tensorflow, Pytorch, etc.).
• Fluency in written and spoken English is essential.
• Communication skills in French are required too.
• A good publication record will be a plus.
The position is open immediately (February 2022). Review of applications will begin as soon as applications are received and continue until the position is filled.

Salaire indicatif
2.591,89 €
Informations de contact

• Victoria Eyharabide (maria-victoria.eyharabide@sorbonne-universite.fr)
• Laurence Likforman-Sulem (Laurence.likforman@telecom-paris.fr)