Recreational walking: Choose your own adventure path
Term
4. term
Education
Publication year
2024
Submitted on
2024-05-29
Pages
101
Abstract
Navigationsapplikationer i dag fokuserer på at optimere ruten ved at finde den hurtigste vej fra punkt A til B, men de er nødvendigvis ikke til at lave rekreative ruter. Derfor har vi udviklet en løsning der kan generere ruter ud fra en persons præfenrener. Det har vi gjort ved at lave vores eget datasæt over Aalborg centrum ud af street view billeder, hvor vi har annoteret 32 forskellige klasser fordelt over 7 overclasser. Vi har brugt overklasserne til at lave en demo med henblik over skræddersyet ruter ud fra kategorier som en person kunne vælge. Derudover, har vi også lavet en machine learning model som kan kategorisere de 7 overklasser med succes over billder i Aalborg. Test- personerne kunne godt lide konceptet bag idéen ved at prøve en tilsvarende virtuel rute, og kunne se dem selv bruge det nye steder som de ikke har været før. Denne forskning er en start på hvordan man kan lave skrædersyet ruter til personer ved hjælp af machine learning, men der mangler at blive lavet en mobilaplikatiion til at bruge idéen i praktis.
Today's navigation applications focus on optimizing the route by finding the fastest way from point A to B, but they are not necessarily designed to create recreational routes. Therefore, we have developed a solution that can generate routes based on a person's preferences. We achieved this by creating our own dataset of central Aalborg using street view images, where we have annotated 32 different classes distributed over 7 superclasses. We used the superclasses to create a demo aimed at custom routes based on categories that a person could choose. Additionally, we have also developed a machine learning model that can successfully categorize the 7 superclasses from images in Aalborg. Test subjects liked the concept behind the idea by trying a similar virtual route and could see themselves using it in new places they had not been before. This research is a start on how to create tailored routes for individuals using machine learning, but a mobile application still needs to be developed to use the idea in practice.
Keywords
Documents