Master’s Thesis: AI-Driven Classification for Road and Infrastructure Issues
Vi svarar vanligtvis inom två veckor
Background
A module within the BM System handles the logistics of reporting and addressing road damages and similar issues, such as broken signs, potholes, tree branches on the road, etc. We refer to these as action points. When a damage is addressed, one or more pictures are taken, and they are manually classified by the person performing the task. There are multiple classes and subclasses, so it can take time for the user to correctly classify the action.
The classes are organised hierarchically, as in the example below, although the exact structure may vary slightly between customers.
When the user has resolved an action point, they take a picture using a BM System app on their mobile phone and classify it on the mobile device, if feasible.
Work Description
Initially, the student should familiarise themselves with the problem and study existing solutions for similar issues. Once this is done, an inventory of the classifications and groupings currently used for action points should be conducted, preferably by looking at multiple customers and then selecting a “test customer” to focus on. Filter the relevant data for the thesis based on its quality and age.
Select an appropriate programming language for the implementation of the classification model and implement it as a proof of concept.
Since internet connectivity is not always available or may be unreliable, it’s preferable to run the trained model locally on the mobile device, which means the model needs to be lightweight.
If this proves impractical, a server-based solution could be considered, where instead of "live classification," the model would alert supervisors if an action point is suspected to have been classified incorrectly. The student is encouraged to propose alternative use cases at any stage of the project, and an open dialogue about the project's scope and methodology is welcomed.
The Student
We would prefer that you have experience with machine learning/AI, either through relevant courses or through personal interest demonstrated by home projects or similar endeavours.
Questions?
Contact Oscar Eriksson, oscare@bmsystem.se or Mikael Östlund, mikaelo@bmsystem.se if you have questions regarding this Master’s Thesis.
- Avdelning
- Student
- Roll
- Examensjobb Master
- Platser
- Uppsala
Uppsala
Arbetsplats och kultur
Vi är ett team med stor kompetens, starkt engagemang och härlig gemenskap. Genom entusiasm och vårt entreprenöriella driv, har vi förmågan att se nya möjligheter och skapa briljanta system. Atmosfären är öppen och vi har en god gemenskap och sammanhållning.
Hjärtat på rätta stället
Vi tar ansvar i hur vår verksamhet påverkar samhället. Vi jobbar aktivt med frågor som rör samhälle, arbetsförhållanden, klimatpåverkan och långsiktig lönsamhet. Våra system bidrar till att minimera körsträckor och bränsleförbrukning samt underlättar för användarna genom att de erhåller konkreta beslutsunderlag från systemen.
Om BM System
BM System är ett Uppsala-baserat mjukvaruföretag som grundades 1991. Verksamheten består av utveckling och marknadsföring av våra egna produkter samt konsultverksamhet och projektledning.
Master’s Thesis: AI-Driven Classification for Road and Infrastructure Issues
Läser in ansökningsformuläret
Jobbar du redan på BM System?
Hjälp till i rekryteringen och hitta din framtida kollega.