Project-Team Texmex
Members
Overall Objectives
Research Program
Application Domains
Software and Platforms
New Results
Bilateral Contracts and Grants with Industry
Partnerships and Cooperations
Dissemination
Bibliography
XML PDF e-pub
PDF e-Pub

previous Previous | up Home | Next next

Section: Software and Platforms

Software

When applicable, we provide the IDDN is the official number, which is obtained when registering the software at the APP (Agence de Protection des Programmes).

New Software
DeCP-Index

Participants : Laurent Amsaleg [Correspondent] , Gylfi Gudmundsson, Diana Moise, Denis Shestakov.

DeCP-Index is a Map-Reduce oriented implementation of the vectorial quantization scheme developped during the PhD of Gylfi Gudmundsson. It is in Java.

First APP deposit: IDDN.FR.001.500011.000.S.P.2013.000.40000

DeCP-Scripts

Participants : Laurent Amsaleg [Correspondent] , Gylfi Gudmundsson, Diana Moise, Denis Shestakov.

DeCP-Scripts is a series of script for installing, configuring and deploying the Map Reduce framework over the grid infrastructure.

First APP deposit: IDDN.FR.001.500012.000.S.P.2013.000.40000

*SVM

Participants : François Poulet [correspondent] , Thanh Nghi Doan.

Soon available from the home page of the Inria Parallel Benchmark Suite http://www.irisa.fr/alf/index.php?option=com_content&view=article&id=82&Itemid=&lang=fr .

*SVM include a set of parallel and incremental SVM classifiers for large scale classification tasks on GPU, CPU or cluster / Grid.

Main software started before 2012
Peyote

Participants : Sébastien Campion, Jonathan Delhumeau [correspondent] , Hervé Jégou.

Peyote is a framework for Video and Image description, indexation and nearest neighbor search. It can be used as-is by a video-search or image-search front-end with the implemented descriptors and search modules. It can also be used via scripting for large-scale experimentation. Finally, thanks to its modularity, it can be used for scientific experimentation on new descriptors or indexation methods. Peyote is used in the Aabot software.

First APP deposit: IDDN.FR.001.4200008.000.S.P.2012.000.20900 .
Aabot

Participant : Jonathan Delhumeau.

Aabot is a tool to facilitate annotation of large video databases. It's primary design focus has been for the annotation on commercials in two 6-month long TV databases. The software keeps a database of already annotated commercials and suggests when it finds a new probable instance. It also validates user annotations by suggesting similar existing commercials if it finds any which are similar by name or content. The user can then confirm the creation of new commercials or accept the correction if he was mistaken.

Aabot is accessed via a web-browser. It is mostly used by uploading and downloading an annotation file. An interactive HTML5 interface is also available when some user feedback is needed (during validation). It uses Peyote as an description / indexing engine.

First APP deposit: IDDN.FR.001.4200010.000.S.P.2012.000.20900 .
Pqcodes

Participant : Hervé Jégou [correspondent] .

Jointly maintained with Matthijs Douze, Inria/LEAR.

Pqcodes is a library which implements the approximate k nearest neighbor search method of  [88] based on product quantization. This software has been transferred to two companies (in August 2011 and May 2012, respectively).

The current version registered at the APP is IDDN.FR.001.220012.001.S.P.2010.000.10000 .

Yael

Participant : Hervé Jégou [correspondent] .

Jointly maintained with Matthijs Douze, from Inria/LEAR.

Yael is a C/python/Matlab library providing (multi-threaded, Blas/Lapack, low level optimization) implementations of computationally demanding functions. In particular, it provides very optimized functions for k-means clustering and exact nearest neighbor search. The library has been downloaded about 2,000 times in 2013.

The current version registered at APP is IDDN.FR.001.220014.001.S.P.2010.000.10000 .
BonzaiBoost

Participant : Christian Raymond [correspondent] .

Available at http://bonzaiboost.gforge.inria.fr/ .

BonzaiBoost stands for boosting over small decisions trees. BonzaiBoost is a general purpose machine-learning program based on decision tree and boosting for building a classifier from text and/or attribute-value data. Currently one configuration of BonzaiBoost is ranked first on http://mlcomp.org a website which propose to compare several classification algorithms on many different datasets

Irisa_Ne

Participant : Christian Raymond [correspondent] .

Irisa_Ne is a couple of Named Entity tagger, one of them is based on CRF and the other HMM. It is dedicated to automatic transcriptions of speech. It does not take into account uppercase or punctuation and has no concept of sentences. However, they also manage texts with punctuation and capitalization.
IRISA News Topic Segmenter (irints)

Participants : Guillaume Gravier [correspondent] , Pascale Sébillot, Anca-Roxana Simon.

This software is dedicated to unsupervised topic segmentation of texts and transcripts. The software implements several of our research methods and is particularly adapted for automatic transcripts. It provides topic segmentation capabilities virtually for any word-based language, with presets for French, English and German. The software has been licensed to several of our industrial partners.

Other softwares

previous Previous | up Home | Next next