Generating natural language descriptions of landscape paintings
Organisatsiooni nimi
Language Technology
Kokkuvõte
Ambitious bachelor students or master students can take this thesis subject. It would be nice (though not required) that the student is interested in art in general and in landscape paintings in particular.
The purpose of the thesis is to generate automatically natural language descriptions of the landscape paintings. The state of the art technology outputs natural language descriptions of the everyday images with reasonable accuracy. Thus, a system can label a photo with a list of tags (e.g., coffee, breakfast, cup) and then generate descriptions like: "A cup of coffee for breakfast to wake up." However, the landscape paintings' descriptions are much more challenging as they involve associating feelings with the picture.
The first stage is to collect natural language descriptions of the landscape paintings. Fortunately, there are many data sources to accomplish this task (Wikipedia, The National Gallery website, etc ...).
The second stage is to generate landscape painting descriptions using various neural network architectures and then find a modality to evaluate the results.
The purpose of the thesis is to generate automatically natural language descriptions of the landscape paintings. The state of the art technology outputs natural language descriptions of the everyday images with reasonable accuracy. Thus, a system can label a photo with a list of tags (e.g., coffee, breakfast, cup) and then generate descriptions like: "A cup of coffee for breakfast to wake up." However, the landscape paintings' descriptions are much more challenging as they involve associating feelings with the picture.
The first stage is to collect natural language descriptions of the landscape paintings. Fortunately, there are many data sources to accomplish this task (Wikipedia, The National Gallery website, etc ...).
The second stage is to generate landscape painting descriptions using various neural network architectures and then find a modality to evaluate the results.
Lõputöö kaitsmise aasta
2020-2021
Juhendaja
Eduard Barbu
Suhtlemiskeel(ed)
inglise keel
Nõuded kandideerijale
Tase
Bakalaureus, Magister
Märksõnad
Kandideerimise kontakt
Nimi
Eduard Barbu
Tel
E-mail