Skip to main content Contact
UE logo
The title of the project

Automatyczna ocena jakości klasyfikatorów obrazu na platformie STIKO

Name of the beneficiary
GDAŃSKIE WYDAWNICTWO OŚWIATOWE SPÓŁKA Z OGRANICZONĄ ODPOWIEDZIALNOŚCIĄ SPÓŁKA KOMANDYTOWA
Project value
484 989,00 zł
Co-financing from the EU
335 155,00 zł
Voivodeship
pomorskie
County
m. Gdańsk
program
Program Operacyjny Inteligentny Rozwój
action
2.3. Proinnowacyjne usługi dla przedsiębiorstw
fund
Europejski Fundusz Rozwoju Regionalnego
perspective
2014 - 2020

SA 42799(2015/X) Proces wytwarzania mobilnych aplikacji inteligentnego lektora (AIL) w Gdańskim Wydawnictwie Oświatowym opiera się na platformach STAIL i STIKO (System Treningu Inteligentnych Klasyfikatorów Obrazu). Kluczowym komponentem aplikacji jest inteligentny klasyfikator obrazu (IKO), w głównej mierze stanowiący głęboką sieć splotową, który jest m.in. produktem treningu prowadzonego na platformie STIKO. Przedmiotem projektu jest wprowadzenie nowych funkcjonalności do podsystemu STIKO oraz optymalizacja procesu wytwarzania AIL. Optymalizacje dotyczą głównie zautomatyzowanej budowy zbioru testowego i walidacyjnego oraz rozbudowy funkcji augmentera. Oba kroki mają na celu podniesienie jakości rozpoznawania obrazów przez IKO oraz pełną automatyzację ich wytwarzania. Zakres projektu obejmuje zatem zaprojektowanie, zaimplementowanie, przetestowanie oprogramowania do: dynamicznego generowania zbioru walidacyjnego oraz wybór modelu i zakończenie treningu w optymalnym momencie na jego podstawie; możliwości przeprowadzenia augmentacji online z wykorzystaniem nowoczesnych architektur procesorów; boosting oraz wyznaczanie metryk “per strona”; inteligentny mechanizm wyznaczania obszarów zainteresowań (ROI) pozwalający na określenie istotności fragmentów strony podczas treningu; oraz wykorzystanie nowatorskiego mechanizmu “transfer testing”. Rezultatem projektu będzie zmodyfikowane oprogramowanie elaSTIKO, pozwalające na wytwarzanie lepszych jakościowo AIL oraz utworzenie nowej usługi B2B. Największą technologiczną innowacją będzie zastosowanie dynamicznych zbiorów walidacyjnych w celu zapewnienia największej jakości klasyfikacji. Aktualnie wejściem dla STIKO jest zbiór o liczbie przykładów równej liczbie stron w książce. Zbiory testowe i walidacyjne tworzone są tylko poprzez mechanizm augmentacji. STIKO ma umożliwiać wytwarzanie IKO bez wymaganej, specjalistycznej wiedzy z zakresu uczenia maszynowego i rozpoznawania obrazów w sposób automatyczny.

Your project description

To add a photo, description, specify a location or attach other materials about the project, you should fill out the form below. After clicking the “send” button, a message will be sent to you e-mail asking for confirmation. You must accept to submit the form. The EU Dot Map Guidelines and consent to the processing of personal data. materials will be sent to the editor of the site, who will publish it on the website after verification test.

Add photos project photos
Add a movie or presentation project films
Description of the project
Opis projektu

I declare that I have read the Regulations of the Online Website of the EU Mapa Dotacji and accept its provisions.

I agree to the processing of my personal data, entered in the contact form, in order to reply to the message. According to art. 13 para. 1 and 2 of the Regulation of the European Parliament and of the Council (EU) 2016/679 of 27 April 2016 on the protection of individuals with regard to the processing of personal data and on the free movement of such data and repealing Directive 95/46 / EC (Journal of Laws EU L 119, 04/05/2016, page 1) (RODO), I inform that:
1.administrator of personal data processed in the Ministry of Development Funds and Regional Policy, based at: ul. Wspólna 2/4 00-926 Warsaw, is the Minister of Development Funds and Regional Policy;
2. personal data will be processed in order to reply to a message, based on the provisions of art. 6 par. 1 lit. a) RODO;
3. personal data may be made available to other authorities entitled under the relevant provisions to obtain data;
4. in connection with the processing of personal data have the right to: – request access to their personal data, rectify or limit processing, – raise objections to the processing of personal data, – withdraw consent to processing at any time, – submit a complaint to the supervisory body – the President of the Office Personal Data Protection;
5. personal data is not subject to automated processing in order to make decisions, including profiling;
6. contact details to the Data Protection Officer in the Ministry of Development Funds and Regional Policy: Data Protection Supervisor, Ministry of Development Funds and Regional Policy, address: ul. Wspólna 2/4 00-926 Warsaw, e-mail address: IOD@miir.gov.pl