ABSTRACT

Online handwritten scripts have become the most important feeding input to smartphones and tablet devices. Several research investigations are still necessary for the online Arabic handwriting recognition field due to some constraints such as the cursiveness of some scripts, the existence of several groups of similar shape characters, etc.

In order to advance the research and to raise the online handwriting recognition performance of online Arabic script, we organize the ASAR2021 Competition on online Arabic Character Recognition. A report on the competition will be published in the proceedings of ASAR 2021. The report will comprise a description of the participating methods, the evaluation protocol, and the final rankings of the participating algorithms. The results of the competition will also be presented in a dedicated session at ASAR 2021.

 

REGISTRATION

Participants need to register in order to participate in the competition. Please, fill up this form

https://forms.gle/qyGZ1RzcyWd9ncRy6 including your information.

The email subject contains "ASAR 2021: OAHCR ".

The email core contains:  name, institution and list of other members of your team.

INSTRUCTIONS

The competition aims to promote innovation in Online Arabic Character Recognition, as well as to provide objective and fair comparisons among methods. The ranking of participants is based on character recognition rate with respect to test samples. Participants will be given a limited time to submit their results after the competition started. The contest will consist of two phases:

• Phase 1: Participants are provided with training data extracted from LMCA Database to train their algorithms.

• Phase 2: Registered participants receive the validation and test datasets during ASAR2021 session.

 

DATA DESCRIPTION

• The training dataset used for this competition consists of 13715 online handwritten characters extracted from LMCA Database (An TXT file including trajectory information),

• The validation dataset contains 1865 samples.

• The test data used for this competition consists of 2151 samples.

• LMCA (Lettres Mots Chiffres Arabes) database contains 30.000 shapes for ten digits, 100.000 shapes for 56 Arabic letters and 500 Arabic words. 55 writers were hosted to participate in the collection of the handwritten LMCA. This database is developed in Research Groups in Intelligent Machines, University of Sfax, Tunisia. For more detail of this dataset, thank you for reading this paper:

 

[1] M.Kherallah and A.Elbaati and H.ElAbed and A.M.Alimi, The On/Off (LMCA) Dual Arabic Handwriting Database, International Conference on Frontiers in Handwriting Recognition, 2008. EVALUATION AND RANKING

 

The objective is to run each Arabic handwritten letter recognizer (trained on the LMCA) on an already published part of the LMCA database and on a new sample not yet published. The recognition results on letter level of each system are compared on the basis of the correct recognized letter. A dictionary can be used and should include all 55 different characters.

 

The system will be tested on a new dataset. Running a Recognizer: We run your recognizer (called myrec) by invoking it from the command line as follows: “myrec dataset.txt  output.txt

„ • dataset.txt: The dataset is just a list of 55 classes of different characters to be recognized. • output.txt: The output file should have one line for each input trace. Each response is given as a pair of values: the class of letter, followed by the confidence.

To guarantee that the description you provide of your method returns the results you have submitted, participants are asked to provide the source code associated to your final submission and provide a detailed description and requirements in a README file. Note that the code will solely use by the Technical Committee to guarantee the fairness of the competition. We will not share this with any third party or make any commercial use of it.

 

IMPORTANT DATES

All deadlines are at 11:59 PM UTC. The competition organizers reserve the right to update the contest timeline if they see it necessary.

• Training: April 5th, 2021.

• Registration: April 5th, 2021 to May 31th, 2021.

• Test sets available: June 1st, 2021. Final submission deadline: June 15th, 2021.