Raphael Baena

Email  /  Scholar  /  Twitter  /  Github

I am a postdoctoral researcher in computer vision at École des Ponts ParisTech (ENPC), under the guidance of Mathieu Aubry. I am part of the Imagine group of the A3SI team, LIGM lab (UMR 8049).

I completed my PhD at IMT Atlantique under the direction of Vincent Gripon and supervision of Lucas Drumetz.

I am interested in computer vision and deep learning, particularly in transfer learning. My current research focuses on applying these techniques to handwritten recognition in historical documents.

profile photo

Research

General Detection-based Text Line Recognition
R. Baena, S. Kalleli, M. Aubry.
Conference: NeurIPS, 2024
Arxiv, Github, Project page
On Transfer in Classification: How Well do Subsets of Classes Generalize?
R. Baena, L. Drumetz, V. Gripon
Conference: EUSIPCO, 2024
Arxiv
Local Mixup: Interpolation of closest input signals to prevent manifold intrusion
R. Baena, L. Drumetz, V. Gripon
Journal Elsevier Signal Processing, 2024
Arxiv , Github
Entropy based feature regularization to improve transferability of deep learning models
R. Baena, L. Drumetz, V. Gripon
Conference: ICASSP, 2023
Arxiv, Github
Inferring Graph Signal Translations as Invariant Transformations for Classification Tasks
R. Baena, L. Drumetz, V. Gripon
Conference: EUSIPCO, 2021
Arxiv

Teaching

Teaching Assistant in Deep Leaning Course Supervisor: Mathieu Aubry
ENPC 2nd Semester 2024
Teaching Assistant in Machine Leaning (15 hrs) Course Supervisor: David Picard
ENPC 1st Semester 2024
Teaching Assistant in Advanced Algorithmics
and Graph Theory with Python
(32 hrs)
Course Supervisor: Vincent Gripon
IMT Atlantique 1st Semester 2022
Supervision of Master's students for research Projects (12 hrs) Course Supervisor: Vincent Gripon
IMT Atlantique 2nd Semester 2021

Reviewer Activity

ICASSP Years: 2023-2024
IEEE Transaction in Signal Processing (TSP) Years: 2022-2023

Seminars / Talks

Interpretable and Principled Historical Document Analysis
Invited Talk at LIRIS, Lyon
29 October 2024
General Detection-based Text Line Recognition
Presentation at the Computer Vision Center (CVC), Barcelona
24 October 2024

Awards

ICDAR 2024 Competition on Handwriting Recognition of Historical Ciphers
Task 2A - Borg and Task 3B - Ramanacoil Cipher
R. Baena, S. Kalleli, M. Aubry
Text Recognition (HTR) in low resource scenarios is a challenging problem. This is particularly the case of historical encrypted manuscripts, so called ciphers. The cipher alphabets oftentimes include digits, Latin or Greek letters, Zodiac and alchemical signs combined with various diacritics, as well as invented symbols.

Website stolen from Jon Barrow.