[visionlist] Call for Participants: IJCB 2023 Competition: 8th Sclera Segmentation and Recognition Benchmarking Competition (SSRBC 2023)

Vitomir Struc vitomir.struc at fe.uni-lj.si
Mon Mar 20 04:09:51 -04 2023


CALL FOR PARTICIPANTS - SSRBC 2023

********************************************************
8th Sclera Segmentation and Recognition  Benchmarking Competition (SSRBC 
2023)

Held in conjunction with IEEE/IAPR IJCB 2023
https://ijcb2023.ieee-biometrics.org/

Important dates: Registration is already open
SSRBC 2023 Website: 
https://sites.google.com/hyderabad.bits-pilani.ac.in/ssrbc2023/home
********************************************************

Sclera biometrics have gained significant popularity among emerging 
ocular traits in the last few years. In order to evaluate the potential 
of this trait, a considerable amount of research has been presented in 
the literature, both employing the sclera individually and in 
combination with the iris. In spite of those initiatives, sclera 
biometrics need to be studied more extensively to ascertain their 
usefulness. Moreover, the sclera segmentation task still requires a 
significant amount of attention due to challenges associated with the 
performance of existing techniques while sclera recognition is performed 
in cross-sensor and resolution scenarios. In order to investigate these 
challenges, document recent development and attract the 
attention/interest of researchers we are planning to host the next 
Sclera Segmentation and Recognition Benchmarking Competition SSRBC 2023. 
SSRBC 2023 will be the 7 th in the series of sclera (segmentation and 
recognition) benchmarking competitions following SSBC 2015, SSRBC 2016, 
SSERBC 2017, SSBC 2018, SSBC 2019 and SSBC 2020 held in conjunction with 
BTAS 2015, ICB 2016, IJCB 2017, ICB 2018, 19 and 20, respectively. Due 
to the overwhelming success of SSBC 2015, SSRBC 2016, SSERBC 2017, SSBC 
2018, 2019 and IJCB 2020, we plan to organize this proposed competition 
to benchmark sclera segmentation and recognition jointly with both 
cross-sensor and low and high-resolution images.

How to participate?

Registration for the competition can be done by email. If you would like 
to register and receive the training dataset, please send an email to 
abhijit.das at hyderabad.bits-pilani.ac.in with the subject line as "SSRBC 
2023 registration" with the following information:

Name, Affiliation, Email, Phone number, CV , Mailing Address and signed 
version of the following form .


Organizers :

Dr. Abhijit Das, BITS Pilani, Hyderabad, India 
(abhijit.das at hyderabad.bits-pilani.ac.in)

Dr. Aritra Mukherjee, BITS Pilani, , Hyderabad, India 
(a.mukherjee at hyderabad.bits-pilani.ac.in)

Prof. Umapada Pal,  Indian Statistical Institute, Kolkata, India 
(umapada at isical.ac.in )

Prof. Peter Peer, University of Ljubljana, Ljubljana, Slovenija 
(peter.peer @fri.uni-lj.si)

Assoc. Prof. Vitomir Štruc , University of Ljubljana, Ljubljana, 
Slovenija (vitomir.struc @fe.uni-lj.si)


Execution

Description of the dataset(s) used for the competition and the available 
annotations

The competition aims to benchmark the sclera segmentation and 
recognition tasks with a dataset containing both low and high-resolution 
images. Three different datasets will be employed for the competition, 
where two were acquired with a DSLR camera and one by a mobile camera.

The first dataset, i.e, the multi-angle sclera dataset (MASD), consists 
of 2624 RGB images taken from 82 identities. Images were collected from 
both the eyes of each individual, so there are 164 different eyes in 
total in the dataset. For each individual image, four gaze directions 
(looking straight, left, right and up) were captured and for each 
direction 4 images were taken. The subjects from the database are both 
male and female and with different eye colors, few of them are wearing 
contact lenses and images were taken at different times of the day. The 
database contains images with blinking eyes, closed eyes and blurred 
eyes. High-resolution images stored in JPEG format are provided in the 
database (7500 x 5000 dimensions). A NIKON D 800 camera and 28300 lenses 
were used for image capturing. A ground truth or manual sclera 
segmentation of this dataset is also available. For development 
purposes, a subset of the database, both eye images and ground truth (1 
image for each angle/gaze of the first 30 subjects, i.e. 120 images in 
total) will be provided to the participants.

The second dataset, the Mobile sclera dataset (MSD), consists of 500 RGB 
images from both eyes of 25 individuals (in other words 50 different 
eyes). For each eye, 10 images were captured. The database contains 
blurred images and images with blinking eyes. The individuals comprise 
both males and females (12 males and 13 females), of different ages and 
different skin colors, 2 of them were wearing contact lenses and the 
images were taken at different times of the day. Variation in image 
quality (blur, lighting condition etc.) and different acquisition 
conditions was included intentionally in the database to investigate the 
performance of the framework in non-ideal scenarios. High-resolution 
images (3264 × 2448) of 96 dpi are included in the database. All the 
images are in JPEG format. The images were captured using a mobile 
camera with an 8-megapixel rear camera.

The third dataset, SBVPI, consists of 1858 RGB images of 110 eyes (i.e., 
55 subjects) captured with a DSLR camera (specifically, a Canon EOS 60D 
with macro lenses). All images were manually cropped to extract the 
desired ROI while maintaining their aspect ratio, then rescaled to 3000 
× 1700 pixels to maintain a consistent image size across the entire 
dataset. Images in the dataset were captured at the highest resolution 
and quality settings available in the camera and in a laboratory 
environment. The dataset contains images taken under 4 different gaze 
directions, with a minimum of 4 images per direction for each subject. 
The appearance variability in SBVPI is due to identity, eye color, 
gender, and age. Manually generated markups of the sclera and periocular 
regions are present for all images. SBVPI is publicly available for 
research purposes.

Details on the experimental protocol and result generation/submission 
procedure,

The competition will address two problems of relevance to IJCB 2023, 
sclera segmentation and recognition, and will be organized around three 
tasks:

● Segmentation task: for the segmentation task, participants will have 
to learn segmentation models on the MASD datasets and then test them on 
the MSD and SBVPI datasets. Complete algorithms will have to be 
submitted for scoring. The final performance evaluation will be 
conducted by the organizers.

● Recognition task: for the recognition task, the participants will be 
asked to develop recognition models on the MASD datasets and then submit 
the trained models for scoring to the organizers. The performance 
evaluation will be conducted on the sequestered MSD and SBVPI dataset. 
In this case, the manually generated (ground truth) segmentation mask 
will be used to get the ROI before subjecting the images to the 
recognition/feature extraction models..

● Joint segmentation and Recognition task: for the joint 
segmentation-recognition task, the participants will be asked to develop 
segmentation as well as recognition models on the MASD datasets and then 
submit the trained models for scoring to the organizers. The performance 
evaluation will be conducted on the sequestered MSD and SBVPI dataset. 
In this case, the segmentation masks generated by the models of the 
participants will be used to extract the ROI. To ensure the models are 
only trained on the vasculature of the sclera, the segmentation masks 
generated by the segmentation models will be used to remove all parts of 
the images that do not belong to the sclera prior to subjecting images 
to the recognition model/feature extractor.

Description of the evaluation criteria (performance metrics) and 
available baseline implementations/code (e.g., a starter kit).

● Segmentation task: The evaluation measures will be precision and 
recall (recall will consider the prior measure for ranking the 
algorithms). The ground truth of the manually segmented sclera region in 
an eye image is constructed, which will be used as a baseline.

● Recognition task: For the recognition task, we will consider 
verification experiments and report the Area Under the ROC Curve (AUC) 
as our main competition metric. For the summary paper, other relevant 
performance indicators will also be reported.


A detailed timeline for the competition:

● Site opens 14th Feb 2023

● Registration starts 14th Feb 2023

● Test dataset available 28th Feb 2023

● Registration closes 10th May 2023

● Algorithm submission deadline 10th May 2023

● Results and report announcement 15th May 2023


Relevant publications

● M. Vitek, A.Das et al., "Exploring Bias in Sclera Segmentation Models: 
A Group Evaluation Approach," in IEEE Transactions on Information 
Forensics and Security, vol. 18, pp. 190-205, 2023, doi: 
10.1109/TIFS.2022.3216468.

● V. Matej, A. Das et al. , SSBC 2020: Sclera Segmentation Benchmarking 
Competition in the Mobile Environment, IJCB 2020.

● A. Das, U Pal, M. Blumenstein, C. Wang, Y. He, Y. Zhu, Z. Sun, Sclera 
Segmentation Benchmarking Competition in Cross-resolution Environment, 
ICB 2019.




More information about the visionlist mailing list