I am happy to announce that the following paper has been accepted to the
2nd International Workshop on Robust Reading (IWRR), in conjunction with ECCV,
that will take place in Amsterdam, on the 9th of September 2016:
From text detection to text segmentation:
a unified evaluation scheme
Stefania Calarasanu (1), Jonathan Fabrizio (1) and Séverine Dubuisson (2)
(1) LRDE-EPITA, 14-16, rue Voltaire, F-94276, Le Kremlin
Bicêtre, France
(2) CNRS, UMR 7222, ISIR, F-75005, Paris, France
Abstract:
Current text segmentation evaluation protocols are often incapable of
properly handling different scenarios (broken/merged/partial characters).
This leads to scores that incorrectly reflect the segmentation accuracy.
In this article we propose a new evaluation scheme that overcomes most
of the existent drawbacks by extending the EvaLTex protocol (initially
designed to evaluate text detection at region level). This new unified platform
has numerous advantages: it is able to evaluate a text understanding system
at every detection stage and granularity level (paragraph/line/word and now
character) by using the same metrics and matching rules; it is robust to all
segmentation scenarios; it provides a qualitative and quantitative evaluation
and a visual score representation that captures the whole behavior of a
segmentation algorithm. Experimental results on nine segmentation algorithms
using different evaluation frameworks are also provided to emphasize the
interest of our method.
Ana Stefania Calarasanu
PhD Engineer - EPITA Research and Development Laboratory (LRDE)
14-16 rue Voltaire, 94276 Le Kremlin-Bicêtre CEDEX, France
https://www.lrde.epita.fr/wiki/User:Calarasanu <https://www.lrde.epita.fr/wiki/User:Calarasanu>
About four hundred commits and five months after Vcsn 2.2, we are proud to
announce the release of Vcsn 2.3, code-named "the tuple release"!
http://vcsn.lrde.epita.fr/Vcsn2.3
As usual, many bugs were fixed (some quite old yet unnoticed so far!).
Noteworthy changes include:
- a particular effort was put on the documentation: there are thirty-five
new documentation notebooks, and about forty others were improved.
- full support for a "tuple" operator on all entities: expressions,
polynomials, automata, etc.
In [13]: aut = lambda e: vcsn.context('lan, q').expression(e).automaton()
In [14]: a = aut('[ab]*') | aut('x')
In [15]: a.shortest(6)
Out[15]: \e|x + a|x + b|x + aa|x + ab|x + ba|x
It is also available in the rational expressions themselves:
In [16]: c = vcsn.context('lat<lan, lan>, q'); c
Out[16]: {...}? x {...}? -> Q
In [17]: e = c.expression('[ab]*|x'); e
Out[17]: (a+b)*|x
In [18]: e.shortest(6)
Out[18]: \e|x + a|x + b|x + aa|x + ab|x + ba|x
The derived-term algorithm supports this operator, and generates
equivalent multitape automata.
- many error messages were improved, to help users understand their
mistakes. For instance, instead of
In [2]: vcsn.Q.expression('a**').derivation('a')
RuntimeError: q: star: invalid value: 1
we now display:
In [2]: vcsn.Q.expression('a**').derivation('a')
RuntimeError: Q: value is not starrable: 1
while computing derivative of: a**
with respect to: a
Besides, longuish stack traces under Python are now filtered from
internal details.
- in addition to `%automaton a`, which allows interactive edition of
automata, the notebooks now feature two new interactive editors:
`%context c` to edit/create context `c`, and `%expression e` for
expressions (with an interactive display of the generated automata).
- one may now generate random rational expressions and control the
operators and their probabilities.
- a lot of code improvement and consistency enforcement, both in C++ and in
Python.
For more details, please, see the news page (see below).
People who worked on this release:
- Akim Demaille
- Clément Gillard
- Lucien Boillod
- Raoul Billion
- Sébastien Piat
- Thibaud Michaud
People who influenced this release:
- Alexandre Duret-Lutz
- Jacques Sakarovitch
- Luca Saiu
- Sylvain Lombardy
Vcsn 2.3: http://vcsn.lrde.epita.fr/Vcsn2.3
News page: http://vcsn.lrde.epita.fr/News_File
Vcsn: http://vcsn.lrde.epita.fr
Playground: http://vcsn-sandbox.lrde.epita.fr
Doc: https://vcsn.lrde.epita.fr/dload/2.3/notebooks/!Read-me-first.html
Bonjour,
Nous avons le plaisir de vous annoncer que le Rapport d’activité
quinquennal du LRDE est désormais disponible sur notre site.
Il retrace les activités du LRDE entre 2011 et 2015.
Vous pouvez le télécharger ici :
http://www.lrde.epita.fr/dload/rapport_activite/ra2015.pdf
--
Daniela Becker
Responsable administrative du LRDE
We are pleased to announce that the following paper was accepted to
the 13th International Colloquium on Theoretical Aspects of Computing
(ICTAC 2016).
Derived-term Automata for Extended Weighted Rational Expressions
Akim Demaille
EPITA/LRDE
http://publis.lrde.epita.fr/demaille.16.ictac
We present an algorithm to build an automaton from a rational
expression. This approach introduces support for extended weighted
expressions. Inspired by derived-term based algorithms, its core
relies on a different construct, rational expansions. We introduce an
inductive algorithm to compute the expansion of an expression from
which the automaton follows. This algorithm is independent of the
size of the alphabet, and actually even supports infinite alphabets.
It can easily be accommodated to generate deterministic (weighted)
automata. These constructs are implemented in Vcsn, a free-software
platform dedicated to weighted automata and rational expressions.
