Institutional Repository of Vilnius University
Semantinė teksto-nuotraukos paieška nekilnojamo turto projektams /
Abstract
dc:descriptionThis bachelor’s thesis explores the application of semantic text-image search methods within the Lithuanian real estate market. The study analyzes the retrieval of real estate projects based on specific visual aspects, such as architectural style or particular architectural features. Different methods of semantic image search are compared to determine their effectiveness in this context. Considering that the primary audience in the Lithuanian real estate market is predominantly Lithuanian, search efficacy is evaluated using both English and Lithuanian search queries. The findings indicate that the performance of semantic search utilizing the CLIP joint text-image embedding models surpasses the performance of semantic search based on text embeddings of image descriptions. The research demonstrates that real estate project search can be effectively conducted based on broad characteristics, such as historical architectural style or prominent features, such as bodies of water. However, the models exhibit limitations in identifying images based on fine-grained details, such as bike paths or roof terraces. Furthermore, it is determined that although semantic search for Lithuanian queries can be implemented using multilingual embedding models, the search efficacy using Lithuanian queries does not match the efficacy of search conducted using English queries.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Bujevičius, Ignas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD210645345&prefLang=en_US