Browsing by Author "Caetano, Alan Nunes"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
- Cyclist route assessment using machine learningPublication . Caetano, Alan Nunes; Lima, Edirlei Soares de; Estima, JacintoIncreasing the number of bike commutes can provide numerous benefits for individuals and communities. However, several factors including the availability of cycle paths, traffic characteristics, and pavement quality, can either encourage or discourage the use of bicycles. To promote cycling and understand how cyclists interact with the urban environment, it is crucial to assess the quality of cyclist routes. This thesis proposes an automatic assessment tool that uses machine learning to detect features of the route segment and calculates a score representing the level of safety and comfort for cyclists. The models are trained on YOLOv5 to classify pavement types, detect pavement defects and detect the presence of cycle paths. Two datasets were built and annotated for the pavement type classification and cycle infrastructure detection tasks. A questionnaire was applied to cyclists to compare the real perceptions with the automatic assessment. The results showed a good alignment with the real perceptions, validating the approach, but also demonstrated the need of adding new features and improving the models’ performance before being adequate for real use.