[image 00586] 講演会のお知らせ: Dr. Herve Jegou, INRIA

[Shin'ichi Satoh]

image-mlの皆様、

国立情報学研究所の佐藤真一です。INRIAのHerve Jegou博士による以下の講演
会のお知らせをお送りいたします。参加無料、参加登録も不要です。どうぞご
参加下さい。

-- Shin'ichi


日時: 5月12日(月) 14:00-15:00
場所: 国立情報学研究所 19階 会議室1901/1902
http://www.nii.ac.jp/access/

Title:
  Reducing the interferences in match kernels for image recognition

Speaker:
  Dr. Herve Jegou, INRIA

Abstract:

We consider the design of a single vector representation for an image 
that embeds and aggregates a set of local patch descriptors such as 
SIFT. More specifically we aim to construct a dense representation, like 
the Fisher Vector or VLAD, though of small or intermediate size.

After introducing the context of match kernels, we will present two 
contributions, both aimed at regularizing the individual contributions 
of the local descriptors in the final representation. The first is a 
novel embedding method that avoids the dependency on absolute distances 
by encoding directions. The second is a "democratization" strategy that 
further limits the interaction of unrelated descriptors in the 
aggregation stage.

This is joint work with Andrew Zisserman.

Short Bio:

Herve Jegou defended his thesis in 2005, on joint source channel coding 
to improve the transmission of compressed data.
He joined Inria Grenoble as a permanent researcher in 2006, in the LEAR 
team, to work on computer vision problems. He moved to INRIA Rennes in 
2009. His research interests include pattern recognition, in particular 
for large collections of images and videos.