Recreational walking: Choose your own adventure path
Authors
Phillips, Jacob William Kragh ; Mikkelsen, Calvin Lloren
Term
4. term
Education
Publication year
2024
Submitted on
2024-05-29
Pages
101
Abstract
De fleste navigationsapps optimerer for den hurtigste vej fra A til B, ikke for behagelige, rekreative ruter. Vi har udviklet et system, der kan generere ruter ud fra en persons præferencer. For at gøre dette byggede vi vores eget datasæt for det centrale Aalborg baseret på street view-billeder. Vi mærkede 32 detaljerede klasser og samlede dem i 7 bredere "superklasser" (overordnede kategorier). Disse superklasser bruges i en demo, hvor brugeren kan vælge kategorier og få skræddersyede ruter. Vi har også trænet en maskinlæringsmodel, der kan genkende de 7 superklasser i billeder fra Aalborg. I en test prøvede deltagere en lignende virtuel rute, kunne lide konceptet og kunne se sig selv bruge det nye steder, de ikke havde været før. Vores arbejde er et første skridt mod personlige ruter med maskinlæring; der mangler stadig at udvikle en mobilapp til praktisk brug.
Most navigation apps optimize for the fastest route from A to B rather than enjoyable, recreational routes. We developed a system that can generate routes based on a person's preferences. To support this, we built our own dataset of central Aalborg using street-view images. We labeled 32 detailed classes and grouped them into 7 broader "superclasses" (higher-level categories). These superclasses power a demo that lets users choose categories to get custom routes. We also trained a machine learning model that can recognize the 7 superclasses in images from Aalborg. In a test, participants tried a similar virtual route, liked the concept, and could imagine using it in places they had not visited before. Our work is an early step toward personalized routes using machine learning; a mobile app still needs to be developed for real-world use.
[This summary has been rewritten with the help of AI based on the project's original abstract]
Keywords
Documents