In order to join the MICHE-II competition, each participant
is asked to submit a registration form first. Once correctly registered, the
participant can also submit one (and only one) executable application
MICHE-II is a challenge for iris recognition on MICHEDB (a dataset of images grabbed
by mobile devices and available online here). Each executable
should therefore be able to receive from command line a pair of images from
the dataset and a pair of corresponding segmentation masks and should produce
a score in terms of dissimilarity between the two irises.
The order of inputs is strictly defined. Let
I1 = image1Filename.ext be the first RGB image containing an iris;
M1 = mask1Filename.ext be the binary mask of I1;
I2 = image2Filename.ext be the second RGB image containing an iris;
M2 = mask2Filename.ext be the binary mask of I2;
path be the directory for matching results;
Let APP be the executable application, then by running:
APP I1 M1 I2 M2 path
a TXT file containing the dissimilarity_score
is created. Such TXT file must have the following properties:
it is saved in path (preferably something like "./results")
its filename is image1Filename_image2Filename.txt (NOTE. filenames without file extensions)
its content is image1Filename [whitespace] image2Filename [whitespace] dissimilarity_score (NOTE. filenames without file extensions)
The dissimilarity score
is meant as the probability that two irises are from two different subjects. The higher is the
dissimilarity the higher is the probability that the two irises are not from
the same person. Let I be set of
images from MICHEDB, the dissimilarity
function D is defined as:
where
and satisfies the following properties:
The participants can use the whole MICHEDB dataset for developing and performing
experimentations of their proposed algorithm. The participants should take into account that the dataset is going to
be extended with new acquisitions by new mobiles and of new subjects according to the same
acquisition protocol applied to the current version of the database.
The challenge will be run on a subset of the new version of MICHEDB that will be revealed
together with the final ranking.
Participants must consider that the best segmentation algorithm submitted to MICHE-I
will be used to generate the binary masks. Since it will be used also for the final ranking of submitted algorithms, participants are invited to use it for testing their proposal.
Given an RGB image in input, the segmentation algorithm gives in output (a) the binary mask; (b) the normalised mask of the iris region; (c) the normalised RGB iris extracted from the image (see figures below):
(a)
(b)
(c)
Participants must consider that the only one segmentation result to be used by the proposed algorithm is the binary mask (a).
The Unsupervised detection of non-iris occlusions (by Haindl et al.) is available for download here. For details of the algorithm, refer to
the article published and available on-line in Elsevier Pattern Recognition Letters Journal.
Each executable is supposed to be self-contained and it will not have access to the
Internet. No any additional download has to be expected to run the application.
The submitted proposal must therefore contain all supporting files (dlls, libraries
and so on) useful to its proper running.
The executable can be written in any programming language and should run on one of
the following operating systems: (1) Windows 7 64/32 bit, (2) Linux Ubuntu 14.04. Code written in Matlab is
also acceptable at condition that it runs on Matlab 2013. In case of any special setting needed for the proper running of the
algorithm, a README file is expected.
Executables that do not match the requirements above could be discarded from the contest
at the discretion of the Evaluating Committee.