I am a french enginner in applied mathematics and Computer Science.
I have in my previous job, worked on PDE (partial differential equation),
and non-linear optimization.


Screenshot:

I am now research engineer in the LEAR team of INRIA.
My job here is to :

• develop a demonstration software to present the know-how of the LEAR team in image retrieval,
• to implement the required method in an efficient way (image filtering, image resampling, affine keypoints, SIFT descriptor, high dimensional indexing, geometric constraints ... )  or  to find  new method  to improve the process.
  
• to help to administrate the machines of the team (under Mandrake-Linux).
  

The demo :

My demo software is a real time image retrieval software.
We have image database (about 400 images, 400x300),
we take with a webcam few images of the scene around us and
we add them to the database.
We can know retrieve in (almost) real time the image of the database
which is the most similar to the current image of the webcam. 

Currently this process (description + retrieval) is done twice a second
for this database/size of image, on  2.8 GHz PC.


Screenshot:

Here are some screenshots of the demo.
(click on the images to see them in full size)

Top Left image is the query image.

Top Right image is the image in the database
which is the most similar to the query.
On this image I draw the points that match
the query image and a rough estimation
of a global affine transform.


Bottom window shows some sample images of the database.

You can see that this retrieval process is
robust to geometric transformation of the image
and occlusion of a part of the image.




































 
back to the main page