PhD position in "image classification on a large scale" - France
Posted: Sat Jul 03, 2010 5:14 pm
The LEAR research group at INRIA Grenoble and the TVPA research group
at the Xerox Research Center Europe in Grenoble are looking for a PhD
student. The candidate will be jointly supervised and will spend time
in both institutions
Annotated data has in the past been viewed as a scarce resource for
image classification. However, this perception is changing as large
amounts of labeled images are becoming available . As a
consequence, the focus is shifting from "how to learn a category from
a single image?"  to "how to handle large quantities of
data?" [5,6,7,8,11,12]. The focus of this PhD will be image
classification on a large scale, i.e. in the case of millions of
images and thousands of classes.
This PhD will improve the accuracy of image classifiers on a large
scale. A possible solution is to use structured output learning for
large hierarchies of semantic classes. This has shown to be successful
for pattern recognition problems of moderate size . However, it is
unclear how it scales to very large numbers of images and classes .
Another possible research line is the use of image segmentation. It
was shown that object segmentation can boost the
accuracy of image classification . However, this comes at a
significant cost which is incompatible with large-scale datasets.
Hence, efficient alternatives to complete segmentation have to be
developed. While the primary application of large scale
classification is image annotation, the availability of a large
number of accurate image classifiers can have other applications, such
as large-scale image retrieval , where semantic information is
* Masters degree (preferably in Computer Science; Mathematics and
Electrical Engineering will also be considered)
* Solid programming skills; the projects involve programming in C/C
++ (and some Matlab)
* Solid mathematics knowledge (especially linear algebra and
* Creative and highly motivated
* Fluent in English, both written and spoken
* Prior knowledge in the areas of computer vision, machine
learning or data mining is a plus (ideally a master thesis in a
As soon as possible
This is a joint project between INRIA Grenoble and XRCE.
The two research centers are just minutes away from each other.
The candidate will be required to spend time in both institutions.
Dr. Cordelia Schmid, sch...@inrialpes.fr
Dr. Florent Perronnin, florent.perron...@xrce.xerox.com
Please send applications via email, including:
* a complete CV
* graduation marks
* topic of your master thesis
* the name and email address of one reference (preferably your
master thesis supervisor)
* if you already have research experience, please include a
publication list and references.
 J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-
Fei. Imagenet: A large-scale hierarchical image database. In
 L. Fei-Fei, R. Fergus, and P. Perona. One-shot learning of
object categories. IEEE PAMI, 28(4), 2006.
 H. Harzallah, F. Jurie, C. Schmid. Combining efficient object
localization and image classification. In ICCV, 2009.
 H. Jegou, M. Douze, and C. Schmid. Hamming embedding
and weak geometric consistency for large scale image search.
In ECCV, 2008.
 Y. Li, D. Crandall, and D. Huttenlocher. Landmark classification
in large-scale image collections. In ICCV, 2009.
 S. Maji and A. Berg. Max-margin additive classifiers for
detection. In ICCV, 2009.
 F. Perronnin, J. Sanchez and Y. Liu. Large-scale image
categorization with explicit data embedding. In CVPR, 2010.
 F. Perronnin, J. Sanchez and T. Mensink. Improving the Fisher
kernel for large-scale image classification. In ECCV, 2010.
 N. Rasiwasia, P. Moreno, and N. Vasconcelos. Bridging the gap:
Query by semantic example. IEEE Trans. on MM, 9(5):923--938, 2007.
 I. Tsochantaridis, T. Hofmann, T. Joachims, and Y. Altun.
Support vector machine learning for interdependent and
structured output spaces. In ICML, 2004.
 A. Vedaldi and A. Zisserman. Efficient additive kernels via
explicit feature maps. In. CVPR, 2010.
 G. Wang, D. Hoiem, and D. Forsyth. Learning image similarity
from flickr groups using stochastic intersection kernel
machines. In ICCV, 2009.