The “Performance Evaluation of Indirect Immunofluorescence Image Analysis Systems” is an initiative jointly organized by the University of Salerno (Italy) and the University of Queensland (Australia) with the support of the Sullivan Nicolaides Pathology (SNP) – Queensland Medical Testing Laboratory, (Australia).

The involvement of the SNP, a relevant laboratory that offers high qualified pathology services for doctors, private hospitals and nursing homes in Queensland northern New South Wales and Darwin, allowed to obtain significant and valuable datasets that will be used for the contest.


  • Brian C. Lovell (University of Queensland, Australia)
  • Gennaro Percannella (University of Salerno, Italy)
  • Mario Vento (University of Salerno, Italy)
  • Arnold Wiliem (University of Queensland, Australia)


Two different recognition tasks are proposed to the participants. Each research team is allowed to apply to both tasks or just one.

Task 1: Cell level Classification

This classification task is a reproposition of the competition on Cells Classification by Fluorescent Image Analysis hosted on last September 2013 by the 20th IEEE International Conference on Image Processing (ICIP 2013). The main aim is to encourage the participation on one hand of new research teams that intend to start with a task that is simpler than the Task 2 described below and on the other hand of  experienced teams that already competed in 2013 and intend to improve their algorithms. The idea is to maintain this task as a permanent competition with the aim of constantly monitoring the advances of the research in this field.

Detailed information about this task are available here.

Task 2: Specimen level Classification

With this task, the scientific community will have to face with a new and more challenging pattern recognition problem: the classification at specimen level, and differently from the previous task in this case no segmentation information is provided.
To this aim it will be used a brand new very large dataset of HEp-2 images collected with the support of the Sullivan Nicolaides Pathology testing laboratory. The dataset will be composed by 8 bits / grey-level specimen images, each one containing a variable number of cells.

Detailed information about this task are available here.


The competition is open to academic, research and industrial institutions. Each team receives, after a registration procedure, a login/password to access the restricted part of the contest web site in order to download the training set.

Participating teams must give their preference on whether they want to appear as anonymous in the publication of the contest results. The results of all the teams with respect to one or both tasks will be included in the final contest report, in anonymous form if requested. With regard to this aspect, it is worth to point out that results published in the contest report cannot be used for advertising or other commercial purposes. The goal of the competition is merely to assess the state-of-the-art.


The application obtaining the highest value of the accuracy in specimen/cell classification over the test set will be declared as the winners.

Up to two winners will be declared during the competition:

  • Task 1: Cell level classification (a winner will be declared only if the performance of the best team will be higher than the performance of the winner of the ICIP 2013 edition of the competition)
  • Task 2: Specimen level classification

The proclamation of the winners will be made during the contest session at ICPR 2014. The contest session will be held as one of the sessions of the I3A Workshop.