sec_prior.tex

\section{Prior work}
\label{prior work}

\begin{verbatim}
Classic features:
-----------------
VLAD
Fisher Vector
Triangulation embedding
Tolias

Enhancements:
-------------
Spatial Pyramid Pooling (SPP)
Hellinger kernel / root SIFT
Dense sampling
E-SVM

New features:
-------------
CNN features
NetVLAD

Hybrid pooling CNN / aggregator features:
-----------------------------------------
VLAD orderless pooling (exploits SPP)
Conv layers + aggregator: Vedaldi, Kulkarni
SIFT + FCs: Perronin
unsup dense: Deep Fisher vector.


Kernel methods
--------------
\end{verbatim}

% \emph{\color{red} paragraph about image search}

% The image retrieval litterature 

% %where the query image is the positive example and the rank of an search result image is the score of the score of this image in the classifier obtained by the E-SVM. 

% The application of kernel methods for SVM has an extensive bibliography. \emph{\color{red} so extensive that I do not know yet what to put here...}

%%% Local Variables:
%%% TeX-master: "main_eccv"
%%% End: